[llvm-branch-commits] [clang] [clang-tools-extra] [libcxx] [llvm] [mlir] [openmp] [CalledValuePropagation] Fix handling of unnamed functions (PR #244846)
Ömer Sinan Ağacan via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Sat Nov 21 09:50:43 PST 2026
https://github.com/osa1 updated https://github.com/llvm/llvm-project/pull/244846
>From 171940db6ad5b7f20e3a120f08bcabcafb255d46 Mon Sep 17 00:00:00 2001
From: cpp-engine <cpp.thread at gmail.com>
Date: Sat, 21 Nov 2026 11:31:02 +0400
Subject: [PATCH 01/35] [X86]: `sub(0, and(X, 1))` -> `ashr(shl(X, SizeInBits -
1), SizeInBits - 1)` (#243478)
In `InstCombine`, expressions of the form `ashr(shl(X, SizeInBits - 1),
SizeInBits - 1)` are canonicalized into the form `sub(0, and(X, 1))`.
The issue with this representation is that the generated code is
suboptimal and requires an additional register. To address this, an
inverse operation has been added for `x86`.
Fixes https://github.com/llvm/llvm-project/issues/208410
---
llvm/lib/Target/X86/X86ISelLowering.cpp | 14 ++++
llvm/test/CodeGen/X86/sar_fold64.ll | 106 +++++++++++++++++++-----
2 files changed, 100 insertions(+), 20 deletions(-)
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 2b3ab72cd9233b..b269f3a4a1ffe4 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -61439,6 +61439,8 @@ static SDValue combineX86CloadCstore(SDNode *N, SelectionDAG &DAG) {
static SDValue combineSub(SDNode *N, SelectionDAG &DAG,
TargetLowering::DAGCombinerInfo &DCI,
const X86Subtarget &Subtarget) {
+ using namespace SDPatternMatch;
+
EVT VT = N->getValueType(0);
SDValue Op0 = N->getOperand(0);
SDValue Op1 = N->getOperand(1);
@@ -61449,6 +61451,18 @@ static SDValue combineSub(SDNode *N, SelectionDAG &DAG,
/*AllowOpaques*/ false);
};
+ // sub(0, and(X, 1)) -> ashr(shl(X, SizeInBits - 1), SizeInBits - 1)
+ if (VT.isVectorOf(MVT::i16) || VT.isVectorOf(MVT::i32) ||
+ VT.isVectorOf(MVT::i64)) {
+ SDValue X;
+ if (sd_match(N, m_Neg(m_OneUse(m_And(m_Value(X), m_One()))))) {
+ SDValue VecConst =
+ DAG.getShiftAmountConstant(VT.getScalarSizeInBits() - 1, VT, DL);
+ SDValue Shl = DAG.getNode(ISD::SHL, DL, VT, X, VecConst);
+ return DAG.getNode(ISD::SRA, DL, VT, Shl, VecConst);
+ }
+ }
+
// X86 can't encode an immediate LHS of a sub. See if we can push the
// negation into a preceding instruction. If the RHS of the sub is a XOR with
// one use and a constant, invert the immediate, saving one register.
diff --git a/llvm/test/CodeGen/X86/sar_fold64.ll b/llvm/test/CodeGen/X86/sar_fold64.ll
index 245af74c238912..ce22eb9f139745 100644
--- a/llvm/test/CodeGen/X86/sar_fold64.ll
+++ b/llvm/test/CodeGen/X86/sar_fold64.ll
@@ -1,7 +1,8 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=CHECK,SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX,AVX512VL
define i32 @shl48sar47(i64 %a) #0 {
; CHECK-LABEL: shl48sar47:
@@ -70,26 +71,15 @@ define i8 @all_sign_bit_ashr(i8 %x) {
define <4 x i32> @all_sign_bit_ashr_vec0(<4 x i32> %x) {
; SSE-LABEL: all_sign_bit_ashr_vec0:
; SSE: # %bb.0:
-; SSE-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; SSE-NEXT: pxor %xmm1, %xmm1
-; SSE-NEXT: psubd %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: pslld $31, %xmm0
+; SSE-NEXT: psrad $31, %xmm0
; SSE-NEXT: retq
;
-; AVX1-LABEL: all_sign_bit_ashr_vec0:
-; AVX1: # %bb.0:
-; AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpsubd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT: retq
-;
-; AVX2-LABEL: all_sign_bit_ashr_vec0:
-; AVX2: # %bb.0:
-; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
-; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpsubd %xmm0, %xmm1, %xmm0
-; AVX2-NEXT: retq
+; AVX-LABEL: all_sign_bit_ashr_vec0:
+; AVX: # %bb.0:
+; AVX-NEXT: vpslld $31, %xmm0, %xmm0
+; AVX-NEXT: vpsrad $31, %xmm0, %xmm0
+; AVX-NEXT: retq
%and = and <4 x i32> %x, <i32 1, i32 1, i32 1 , i32 1>
%neg = sub <4 x i32> zeroinitializer, %and
%sar = ashr <4 x i32> %neg, <i32 1, i32 31, i32 5, i32 0>
@@ -121,6 +111,14 @@ define <4 x i32> @all_sign_bit_ashr_vec1(<4 x i32> %x) {
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpsubd %xmm0, %xmm1, %xmm0
; AVX2-NEXT: retq
+;
+; AVX512VL-LABEL: all_sign_bit_ashr_vec1:
+; AVX512VL: # %bb.0:
+; AVX512VL-NEXT: vpbroadcastd %xmm0, %xmm0
+; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
+; AVX512VL-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT: vpsubd %xmm0, %xmm1, %xmm0
+; AVX512VL-NEXT: retq
%and = and <4 x i32> %x, <i32 1, i32 1, i32 1 , i32 1>
%sub = sub <4 x i32> <i32 0, i32 1, i32 2, i32 3>, %and
%shf = shufflevector <4 x i32> %sub, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -150,6 +148,13 @@ define <4 x i32> @all_sign_bit_ashr_vec2(<4 x i32> %x) {
; AVX2-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
+;
+; AVX512VL-LABEL: all_sign_bit_ashr_vec2:
+; AVX512VL: # %bb.0:
+; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
+; AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT: vpaddd %xmm1, %xmm0, %xmm0
+; AVX512VL-NEXT: retq
%and = and <4 x i32> %x, <i32 1, i32 1, i32 1 , i32 1>
%add = add <4 x i32> %and, <i32 -1, i32 -1, i32 -1, i32 -1>
%sar = ashr <4 x i32> %add, <i32 1, i32 31, i32 5, i32 0>
@@ -181,6 +186,14 @@ define <4 x i32> @all_sign_bit_ashr_vec3(<4 x i32> %x) {
; AVX2-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpaddd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
+;
+; AVX512VL-LABEL: all_sign_bit_ashr_vec3:
+; AVX512VL: # %bb.0:
+; AVX512VL-NEXT: vpbroadcastd %xmm0, %xmm0
+; AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
+; AVX512VL-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512VL-NEXT: vpaddd %xmm1, %xmm0, %xmm0
+; AVX512VL-NEXT: retq
%and = and <4 x i32> %x, <i32 1, i32 1, i32 1 , i32 1>
%add = add <4 x i32> %and, <i32 -1, i32 1, i32 2, i32 3>
%shf = shufflevector <4 x i32> %add, <4 x i32> undef, <4 x i32> zeroinitializer
@@ -188,4 +201,57 @@ define <4 x i32> @all_sign_bit_ashr_vec3(<4 x i32> %x) {
ret <4 x i32> %sar
}
+define <2 x i64> @all_sign_bit_ashr_vec4(<2 x i64> %x) {
+; SSE-LABEL: all_sign_bit_ashr_vec4:
+; SSE: # %bb.0: # %entry
+; SSE-NEXT: psllw $15, %xmm0
+; SSE-NEXT: psraw $15, %xmm0
+; SSE-NEXT: retq
+;
+; AVX-LABEL: all_sign_bit_ashr_vec4:
+; AVX: # %bb.0: # %entry
+; AVX-NEXT: vpsllw $15, %xmm0, %xmm0
+; AVX-NEXT: vpsraw $15, %xmm0, %xmm0
+; AVX-NEXT: retq
+entry:
+ %bc1 = bitcast <2 x i64> %x to <8 x i16>
+ %and = and <8 x i16> %bc1, splat (i16 1)
+ %sub = sub nsw <8 x i16> zeroinitializer, %and
+ %bc2 = bitcast <8 x i16> %sub to <2 x i64>
+ ret <2 x i64> %bc2
+}
+
+define <2 x i64> @all_sign_bit_ashr_vec5(<2 x i64> %x) {
+; SSE-LABEL: all_sign_bit_ashr_vec5:
+; SSE: # %bb.0: # %entry
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; SSE-NEXT: pslld $31, %xmm0
+; SSE-NEXT: psrad $31, %xmm0
+; SSE-NEXT: retq
+;
+; AVX1-LABEL: all_sign_bit_ashr_vec5:
+; AVX1: # %bb.0: # %entry
+; AVX1-NEXT: vpsllq $63, %xmm0, %xmm0
+; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT: retq
+;
+; AVX2-LABEL: all_sign_bit_ashr_vec5:
+; AVX2: # %bb.0: # %entry
+; AVX2-NEXT: vpsllq $63, %xmm0, %xmm0
+; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT: vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT: retq
+;
+; AVX512VL-LABEL: all_sign_bit_ashr_vec5:
+; AVX512VL: # %bb.0: # %entry
+; AVX512VL-NEXT: vpsllq $63, %xmm0, %xmm0
+; AVX512VL-NEXT: vpsraq $63, %xmm0, %xmm0
+; AVX512VL-NEXT: retq
+entry:
+ %and = and <2 x i64> %x, splat (i64 1)
+ %sub = sub nsw <2 x i64> zeroinitializer, %and
+ ret <2 x i64> %sub
+}
+
attributes #0 = { nounwind }
>From aab5b2033768c13c0753538df9f8940ada9d8b45 Mon Sep 17 00:00:00 2001
From: woruyu <1214539920 at qq.com>
Date: Sat, 21 Nov 2026 16:16:43 +0800
Subject: [PATCH 02/35] [SelectionDAG] Use the widened type to select the shift
amount type (#244781)
This fixes https://github.com/llvm/llvm-project/issues/244518.
CombineZExtLogicopShiftLoad reuses the original shift amount when
widening a shift. On X86, widening an i16 shift to i512 leaves the count
as i8, which cannot represent all valid shift amounts for i512. Use
getShiftAmountOperand to convert the count to the type required by the
widened shift.
Assisted-by: OpenAI Codex
---
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 4 +-
.../CodeGen/X86/zext-logicop-shift-load.ll | 47 +++++++++++++++++++
2 files changed, 49 insertions(+), 2 deletions(-)
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 50781b2d4aa23c..92be269ae479cf 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -15686,8 +15686,8 @@ SDValue DAGCombiner::CombineZExtLogicopShiftLoad(SDNode *N) {
Load->getMemoryVT(), Load->getMemOperand());
SDLoc DL1(N1);
- SDValue Shift = DAG.getNode(N1.getOpcode(), DL1, VT, ExtLoad,
- N1.getOperand(1));
+ SDValue ShiftAmt = DAG.getShiftAmountOperand(VT, N1.getOperand(1));
+ SDValue Shift = DAG.getNode(N1.getOpcode(), DL1, VT, ExtLoad, ShiftAmt);
APInt Mask = N0.getConstantOperandAPInt(1).zext(VT.getSizeInBits());
SDLoc DL0(N0);
diff --git a/llvm/test/CodeGen/X86/zext-logicop-shift-load.ll b/llvm/test/CodeGen/X86/zext-logicop-shift-load.ll
index 3f64a383abd2c9..63e1fe767cc41f 100644
--- a/llvm/test/CodeGen/X86/zext-logicop-shift-load.ll
+++ b/llvm/test/CodeGen/X86/zext-logicop-shift-load.ll
@@ -177,3 +177,50 @@ entry:
ret i64 %1
}
+; The widened shift needs a count type that can represent 511.
+define void @zext_bitfield_to_i512(ptr %out, ptr %data) {
+; X86-LABEL: zext_bitfield_to_i512:
+; X86: # %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movzwl (%ecx), %ecx
+; X86-NEXT: shrl $8, %ecx
+; X86-NEXT: andl $1, %ecx
+; X86-NEXT: movl %ecx, (%eax)
+; X86-NEXT: movl $0, 60(%eax)
+; X86-NEXT: movl $0, 56(%eax)
+; X86-NEXT: movl $0, 52(%eax)
+; X86-NEXT: movl $0, 48(%eax)
+; X86-NEXT: movl $0, 44(%eax)
+; X86-NEXT: movl $0, 40(%eax)
+; X86-NEXT: movl $0, 36(%eax)
+; X86-NEXT: movl $0, 32(%eax)
+; X86-NEXT: movl $0, 28(%eax)
+; X86-NEXT: movl $0, 24(%eax)
+; X86-NEXT: movl $0, 20(%eax)
+; X86-NEXT: movl $0, 16(%eax)
+; X86-NEXT: movl $0, 12(%eax)
+; X86-NEXT: movl $0, 8(%eax)
+; X86-NEXT: movl $0, 4(%eax)
+; X86-NEXT: retl
+;
+; X64-LABEL: zext_bitfield_to_i512:
+; X64: # %bb.0:
+; X64-NEXT: movzbl 1(%rsi), %eax
+; X64-NEXT: andl $1, %eax
+; X64-NEXT: movq %rax, (%rdi)
+; X64-NEXT: movq $0, 56(%rdi)
+; X64-NEXT: movq $0, 48(%rdi)
+; X64-NEXT: movq $0, 40(%rdi)
+; X64-NEXT: movq $0, 32(%rdi)
+; X64-NEXT: movq $0, 24(%rdi)
+; X64-NEXT: movq $0, 16(%rdi)
+; X64-NEXT: movq $0, 8(%rdi)
+; X64-NEXT: retq
+ %val = load i16, ptr %data, align 4
+ %shifted = lshr i16 %val, 8
+ %bit = and i16 %shifted, 1
+ %wide = zext i16 %bit to i512
+ store i512 %wide, ptr %out, align 8
+ ret void
+}
>From 111796e7444bd1ccc72be36186a95576b9930b0f Mon Sep 17 00:00:00 2001
From: David Green <david.green at arm.com>
Date: Sat, 21 Nov 2026 10:16:47 +0000
Subject: [PATCH 03/35] [ARM] Promote basic operations with +bf16 (#244862)
This promotes basic fadd, fsub, fmul and fdiv operation with +bf16,
ensuring we convert them to fp32. FMA is apparently incorrect to promote
in the same way as it double rounds.
---
llvm/lib/Target/ARM/ARMISelLowering.cpp | 3 +
llvm/test/CodeGen/ARM/bf16-instructions.ll | 437 ++++++++++++++++--
llvm/test/CodeGen/Thumb2/bf16-instructions.ll | 2 -
3 files changed, 400 insertions(+), 42 deletions(-)
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 366d33711e8e56..c178e740de3d36 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -583,6 +583,9 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM_,
setOperationAction(ISD::STRICT_FP_ROUND, MVT::bf16, Custom);
setOperationAction(ISD::BF16_TO_FP, MVT::f32, Expand);
setOperationAction(ISD::BF16_TO_FP, MVT::f64, Expand);
+
+ for (auto Op : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV})
+ setOperationAction(Op, MVT::bf16, Promote);
} else {
setOperationAction(ISD::BF16_TO_FP, MVT::f32, Expand);
setOperationAction(ISD::BF16_TO_FP, MVT::f64, Expand);
diff --git a/llvm/test/CodeGen/ARM/bf16-instructions.ll b/llvm/test/CodeGen/ARM/bf16-instructions.ll
index 29fd9cd62721e9..43892c40118269 100644
--- a/llvm/test/CodeGen/ARM/bf16-instructions.ll
+++ b/llvm/test/CodeGen/ARM/bf16-instructions.ll
@@ -3,36 +3,277 @@
; RUN: llc < %s -mtriple armv8a-none-eabihf -mattr=+bf16 | FileCheck %s --check-prefixes=CHECK,CHECK-BF16,CHECK-NOFP16
; RUN: llc < %s -mtriple armv8a-none-eabihf -mattr=+bf16,+fullfp16,+fp64 | FileCheck %s --check-prefixes=CHECK,CHECK-BF16,CHECK-FP16
-;define bfloat @test_fadd(bfloat %a, bfloat %b) #0 {
-; %r = fadd bfloat %a, %b
-; ret bfloat %r
-;}
+define bfloat @test_fadd(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_fadd:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vadd.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fadd:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fadd:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %r = fadd bfloat %a, %b
+ ret bfloat %r
+}
-;define bfloat @test_fadd_fast(bfloat %a, bfloat %b) #0 {
-; %r = fadd fast bfloat %a, %b
-; ret bfloat %r
-;}
+define bfloat @test_fadd_fast(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_fadd_fast:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vadd.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fadd_fast:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fadd_fast:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %r = fadd fast bfloat %a, %b
+ ret bfloat %r
+}
-;define bfloat @test_fsub(bfloat %a, bfloat %b) #0 {
-; %r = fsub bfloat %a, %b
-; ret bfloat %r
-;}
+define bfloat @test_fsub(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_fsub:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vsub.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fsub:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vsub.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fsub:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vsub.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %r = fsub bfloat %a, %b
+ ret bfloat %r
+}
-;define bfloat @test_fmul(bfloat %a, bfloat %b) #0 {
-; %r = fmul bfloat %a, %b
-; ret bfloat %r
-;}
+define bfloat @test_fmul(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_fmul:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vmul.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fmul:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vmul.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fmul:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vmul.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %r = fmul bfloat %a, %b
+ ret bfloat %r
+}
-;define bfloat @test_fmadd(bfloat %a, bfloat %b, bfloat %c) #0 {
-; %mul = fmul fast bfloat %a, %b
-; %r = fadd fast bfloat %mul, %c
-; ret bfloat %r
-;}
+define bfloat @test_fmadd(bfloat %a, bfloat %b, bfloat %c) #0 {
+; CHECK-CVT-LABEL: test_fmadd:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: vmov r1, s1
+; CHECK-CVT-NEXT: vmov r2, s2
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: lsl r1, r1, #16
+; CHECK-CVT-NEXT: vmov s4, r0
+; CHECK-CVT-NEXT: lsl r0, r2, #16
+; CHECK-CVT-NEXT: vmov s2, r1
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vfma.f32 s0, s4, s2
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fmadd:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s4, r0
+; CHECK-NOFP16-NEXT: vmov r1, s2
+; CHECK-NOFP16-NEXT: vmul.f32 s0, s4, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: vmov r0, s0
+; CHECK-NOFP16-NEXT: lsl r1, r1, #16
+; CHECK-NOFP16-NEXT: vmov s0, r1
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fmadd:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s4, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vmov r1, s2
+; CHECK-FP16-NEXT: vmul.f32 s0, s4, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %mul = fmul fast bfloat %a, %b
+ %r = fadd fast bfloat %mul, %c
+ ret bfloat %r
+}
-;define bfloat @test_fdiv(bfloat %a, bfloat %b) #0 {
-; %r = fdiv bfloat %a, %b
-; ret bfloat %r
-;}
+define bfloat @test_fdiv(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_fdiv:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vdiv.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fdiv:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vdiv.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fdiv:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vdiv.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %r = fdiv bfloat %a, %b
+ ret bfloat %r
+}
;define bfloat @test_frem(bfloat %a, bfloat %b) #0 {
; %r = frem bfloat %a, %b
@@ -314,7 +555,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-CVT-NEXT: mov r4, r0
; CHECK-CVT-NEXT: ldrh r0, [r0]
; CHECK-CVT-NEXT: vmov s18, r0
-; CHECK-CVT-NEXT: .LBB5_1: @ %loop
+; CHECK-CVT-NEXT: .LBB11_1: @ %loop
; CHECK-CVT-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-CVT-NEXT: ldrh r0, [r4]
; CHECK-CVT-NEXT: vmov.f32 s16, s18
@@ -322,7 +563,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-CVT-NEXT: mov r0, r4
; CHECK-CVT-NEXT: bl test_dummy
; CHECK-CVT-NEXT: tst r0, #1
-; CHECK-CVT-NEXT: bne .LBB5_1
+; CHECK-CVT-NEXT: bne .LBB11_1
; CHECK-CVT-NEXT: @ %bb.2: @ %return
; CHECK-CVT-NEXT: vmov.f32 s0, s16
; CHECK-CVT-NEXT: vpop {d8, d9}
@@ -337,7 +578,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-NOFP16-NEXT: mov r4, r0
; CHECK-NOFP16-NEXT: ldrh r0, [r0]
; CHECK-NOFP16-NEXT: vmov s18, r0
-; CHECK-NOFP16-NEXT: .LBB5_1: @ %loop
+; CHECK-NOFP16-NEXT: .LBB11_1: @ %loop
; CHECK-NOFP16-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NOFP16-NEXT: ldrh r0, [r4]
; CHECK-NOFP16-NEXT: vmov.f32 s16, s18
@@ -345,7 +586,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-NOFP16-NEXT: mov r0, r4
; CHECK-NOFP16-NEXT: bl test_dummy
; CHECK-NOFP16-NEXT: tst r0, #1
-; CHECK-NOFP16-NEXT: bne .LBB5_1
+; CHECK-NOFP16-NEXT: bne .LBB11_1
; CHECK-NOFP16-NEXT: @ %bb.2: @ %return
; CHECK-NOFP16-NEXT: vmov.f32 s0, s16
; CHECK-NOFP16-NEXT: vpop {d8, d9}
@@ -359,14 +600,14 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-FP16-NEXT: vpush {d8, d9}
; CHECK-FP16-NEXT: vldr.16 s18, [r0]
; CHECK-FP16-NEXT: mov r4, r0
-; CHECK-FP16-NEXT: .LBB5_1: @ %loop
+; CHECK-FP16-NEXT: .LBB11_1: @ %loop
; CHECK-FP16-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-FP16-NEXT: mov r0, r4
; CHECK-FP16-NEXT: vmov.f32 s16, s18
; CHECK-FP16-NEXT: vldr.16 s18, [r4]
; CHECK-FP16-NEXT: bl test_dummy
; CHECK-FP16-NEXT: tst r0, #1
-; CHECK-FP16-NEXT: bne .LBB5_1
+; CHECK-FP16-NEXT: bne .LBB11_1
; CHECK-FP16-NEXT: @ %bb.2: @ %return
; CHECK-FP16-NEXT: vmov.f32 s0, s16
; CHECK-FP16-NEXT: vpop {d8, d9}
@@ -909,11 +1150,62 @@ define bfloat @test_copysign_f64(bfloat %a, double %b) #0 {
; away the (fpext (fp_round <result>)) here.
-;define float @test_copysign_extended(bfloat %a, bfloat %b) #0 {
-; %r = call bfloat @llvm.copysign.bf16(bfloat %a, bfloat %b)
-; %xr = fpext bfloat %r to float
-; ret float %xr
-;}
+define float @test_copysign_extended(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_copysign_extended:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: vmov r1, s1
+; CHECK-CVT-NEXT: and r1, r1, #32768
+; CHECK-CVT-NEXT: bfc r0, #15, #17
+; CHECK-CVT-NEXT: orr r0, r0, r1
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bx lr
+;
+; CHECK-NOFP16-LABEL: test_copysign_extended:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: .pad #8
+; CHECK-NOFP16-NEXT: sub sp, sp, #8
+; CHECK-NOFP16-NEXT: vmov r1, s1
+; CHECK-NOFP16-NEXT: vmov r0, s0
+; CHECK-NOFP16-NEXT: uxth r1, r1
+; CHECK-NOFP16-NEXT: strh r1, [sp, #4]
+; CHECK-NOFP16-NEXT: uxth r0, r0
+; CHECK-NOFP16-NEXT: ldrb r1, [sp, #5]
+; CHECK-NOFP16-NEXT: strh r0, [sp]
+; CHECK-NOFP16-NEXT: ldrb r0, [sp, #1]
+; CHECK-NOFP16-NEXT: and r1, r1, #128
+; CHECK-NOFP16-NEXT: and r0, r0, #127
+; CHECK-NOFP16-NEXT: orr r0, r0, r1
+; CHECK-NOFP16-NEXT: strb r0, [sp, #1]
+; CHECK-NOFP16-NEXT: ldrh r0, [sp]
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: add sp, sp, #8
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_copysign_extended:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: .pad #8
+; CHECK-FP16-NEXT: sub sp, sp, #8
+; CHECK-FP16-NEXT: vstr.16 s1, [sp, #4]
+; CHECK-FP16-NEXT: vstr.16 s0, [sp]
+; CHECK-FP16-NEXT: ldrb r0, [sp, #5]
+; CHECK-FP16-NEXT: ldrb r1, [sp, #1]
+; CHECK-FP16-NEXT: and r0, r0, #128
+; CHECK-FP16-NEXT: and r1, r1, #127
+; CHECK-FP16-NEXT: orr r0, r1, r0
+; CHECK-FP16-NEXT: strb r0, [sp, #1]
+; CHECK-FP16-NEXT: vldr.16 s0, [sp]
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: vmov s0, r0
+; CHECK-FP16-NEXT: add sp, sp, #8
+; CHECK-FP16-NEXT: bx lr
+ %r = call bfloat @llvm.copysign.bf16(bfloat %a, bfloat %b)
+ %xr = fpext bfloat %r to float
+ ret float %xr
+}
;define bfloat @test_floor(bfloat %a) #0 {
; %r = call bfloat @llvm.floor.bf16(bfloat %a)
@@ -950,9 +1242,74 @@ define bfloat @test_copysign_f64(bfloat %a, double %b) #0 {
; ret bfloat %r
;}
-;define bfloat @test_fmuladd(bfloat %a, bfloat %b, bfloat %c) #0 {
-; %r = call bfloat @llvm.fmuladd.bf16(bfloat %a, bfloat %b, bfloat %c)
-; ret bfloat %r
-;}
+define bfloat @test_fmuladd(bfloat %a, bfloat %b, bfloat %c) #0 {
+; CHECK-CVT-LABEL: test_fmuladd:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: .vsave {d8}
+; CHECK-CVT-NEXT: vpush {d8}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: vmov.f32 s16, s2
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vmul.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: vmov r0, s16
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: lsl r0, r1, #16
+; CHECK-CVT-NEXT: vmov s2, r0
+; CHECK-CVT-NEXT: vadd.f32 s0, s2, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: vpop {d8}
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-NOFP16-LABEL: test_fmuladd:
+; CHECK-NOFP16: @ %bb.0:
+; CHECK-NOFP16-NEXT: vmov r0, s1
+; CHECK-NOFP16-NEXT: vmov r1, s0
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s0, r0
+; CHECK-NOFP16-NEXT: lsl r0, r1, #16
+; CHECK-NOFP16-NEXT: vmov s4, r0
+; CHECK-NOFP16-NEXT: vmov r1, s2
+; CHECK-NOFP16-NEXT: vmul.f32 s0, s4, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: vmov r0, s0
+; CHECK-NOFP16-NEXT: lsl r1, r1, #16
+; CHECK-NOFP16-NEXT: vmov s0, r1
+; CHECK-NOFP16-NEXT: lsl r0, r0, #16
+; CHECK-NOFP16-NEXT: vmov s2, r0
+; CHECK-NOFP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-NOFP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-NOFP16-NEXT: bx lr
+;
+; CHECK-FP16-LABEL: test_fmuladd:
+; CHECK-FP16: @ %bb.0:
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: vmov r1, s1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s4, r0
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: vmov r1, s2
+; CHECK-FP16-NEXT: vmul.f32 s0, s4, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: vmov r0, s0
+; CHECK-FP16-NEXT: lsl r1, r1, #16
+; CHECK-FP16-NEXT: vmov s0, r1
+; CHECK-FP16-NEXT: lsl r0, r0, #16
+; CHECK-FP16-NEXT: vmov s2, r0
+; CHECK-FP16-NEXT: vadd.f32 s0, s2, s0
+; CHECK-FP16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-FP16-NEXT: bx lr
+ %r = call bfloat @llvm.fmuladd.bf16(bfloat %a, bfloat %b, bfloat %c)
+ ret bfloat %r
+}
attributes #0 = { nounwind }
diff --git a/llvm/test/CodeGen/Thumb2/bf16-instructions.ll b/llvm/test/CodeGen/Thumb2/bf16-instructions.ll
index 2d18a234e94fe6..ec0f13a8a8b9b7 100644
--- a/llvm/test/CodeGen/Thumb2/bf16-instructions.ll
+++ b/llvm/test/CodeGen/Thumb2/bf16-instructions.ll
@@ -2099,8 +2099,6 @@ define bfloat @test_copysign_f64(bfloat %a, double %b) {
ret bfloat %r
}
-; away the (fpext (fp_round <result>)) here.
-
define float @test_copysign_extended(bfloat %a, bfloat %b) {
; CHECK-NOFP-LABEL: test_copysign_extended:
; CHECK-NOFP: @ %bb.0:
>From bacd8db87f307a4829c8f2d273b5787ab7eb8010 Mon Sep 17 00:00:00 2001
From: Brad Smith <brad at comstyle.com>
Date: Sat, 21 Nov 2026 05:23:00 -0500
Subject: [PATCH 04/35] [OpenMP] Don't over-allocate stack in s390x
__kmp_invoke_microtask (#244475)
Five arguments are passed in registers, so max(0, argc - 3) of the
p_argv entries go on the stack, as the comment says. The code reserved
max(0, argc - 2) slots, one more than needed.
Co-authored-by: Claude Opus 5.5 <noreply at anthropic.com>
---
openmp/runtime/src/z_Linux_asm.S | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/openmp/runtime/src/z_Linux_asm.S b/openmp/runtime/src/z_Linux_asm.S
index 142df5c4feff86..86e96cfc98747f 100644
--- a/openmp/runtime/src/z_Linux_asm.S
+++ b/openmp/runtime/src/z_Linux_asm.S
@@ -2529,7 +2529,7 @@ __kmp_invoke_microtask:
// The total number of bytes is then max(0, argc - 3)*8 + 8
lgr %r10,%r5
- aghi %r10,-2
+ aghi %r10,-3
jnm 0f
lghi %r10,0
0:
>From 22e9d8a5d88722b3df1f682a28c9796f29b8d071 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Bal=C3=A1zs=20Benics?= <benicsbalazs at gmail.com>
Date: Sat, 21 Nov 2026 10:32:52 +0000
Subject: [PATCH 05/35] [clang][SSAF] Move lit tests to
clang/test/ScalableStaticAnalysis (#241883)
SSAF is not part of the static analyzer proper, so its tests do not
belong under clang/test/Analysis.
This should harmonize the test dir name with the lib and include dir
names.
https://discourse.llvm.org/t/untangle-static-analyzer-and-ssaf-tests/91835/19
- git mv clang/test/Analysis/Scalable clang/test/ScalableStaticAnalysis
- Update the plugin add_subdirectory path and the PR labeler glob
- Add check-clang-scalablestaticanalysis-unit to run
ClangScalableAnalysisTests
How to run SSAF tests from now on:
- Lit tests: ninja check-clang-scalablestaticanalysis (auto-generated
from the directory name, so no explicit target)
- A single lit subdirectory: ninja
check-clang-scalablestaticanalysis-ssaf-linker
- Unit tests: ninja check-clang-scalablestaticanalysis-unit
Authored-By: Claude
---
.github/new-prs-labeler.yml | 2 +-
clang/test/CMakeLists.txt | 2 +-
.../PointerFlow/Inputs/tu-summary-bad-array.json | 0
.../PointerFlow/Inputs/tu-summary-bad-summary.json | 0
.../PointerFlow/Inputs/tu-summary-no-key.json | 0
.../PointerFlow/Inputs/tu-summary.json | 0
.../PointerFlow/Inputs/wpa-result-bad-edges.json | 0
.../PointerFlow/Inputs/wpa-result-bad-id.json | 0
.../PointerFlow/Inputs/wpa-result-empty.json | 0
.../PointerFlow/Inputs/wpa-result-no-key.json | 0
.../PointerFlow/Inputs/wpa-result-odd-count.json | 0
.../PointerFlow/Inputs/wpa-result.json | 0
.../PointerFlow/entity-name-no-conflict.cpp | 0
.../external-inline-function-in-multi-tu.test | 0
.../PointerFlow/lref-to-rref-cast.test | 0
.../PointerFlow/multi-decl-contributor.cpp | 0
.../PointerFlow/multi-dim-pointer-flow-constraint.test | 0
.../PointerFlow/system-header-opt-out.cpp | 0
.../PointerFlow/tu-summary-serialization.test | 0
.../PointerFlow/unsafe-buffer-reachable-cast-cycle.test | 0
.../PointerFlow/wpa-result-serialization.test | 0
.../Inputs/tu-summary-bad-id.json | 0
.../Inputs/tu-summary-no-key.json | 0
.../TypeConstrainedPointers/Inputs/tu-summary.json | 0
.../Inputs/wpa-result-bad-id.json | 0
.../TypeConstrainedPointers/Inputs/wpa-result-empty.json | 0
.../Inputs/wpa-result-no-key.json | 0
.../tu-summary-serialization.test | 0
.../type-constrained-pointers.cpp | 0
...e-buffer-reachable-excludes-type-constrained-main.cpp | 0
...er-reachable-excludes-type-constrained-new-delete.cpp | 0
.../wpa-result-serialization.test | 0
.../UnsafeBufferUsage/Inputs/tu-summary-bad-element.json | 0
.../Inputs/tu-summary-bad-ptr-level.json | 0
.../UnsafeBufferUsage/Inputs/tu-summary-no-key.json | 0
.../UnsafeBufferUsage/Inputs/tu-summary.json | 0
.../UnsafeBufferUsage/Inputs/wpa-result-bad-epls.json | 0
.../UnsafeBufferUsage/Inputs/wpa-result-bad-id.json | 0
.../UnsafeBufferUsage/Inputs/wpa-result-empty.json | 0
.../UnsafeBufferUsage/Inputs/wpa-result-no-key.json | 0
.../UnsafeBufferUsage/Inputs/wpa-result-odd-count.json | 0
.../UnsafeBufferUsage/Inputs/wpa-result.json | 0
.../UnsafeBufferUsage/tu-summary-serialization.test | 0
.../UnsafeBufferUsage/wpa-result-serialization.test | 0
.../call-graph-local-entities.cpp | 0
.../Scalable => ScalableStaticAnalysis}/call-graph.cpp | 0
.../cli-errors-compilation-unit-id.cpp | 0
.../command-line-interface.cpp | 0
.../downgradable-errors.cpp | 0
.../extraction-works-alongside-compilation.cpp | 0
.../Scalable => ScalableStaticAnalysis}/help.cpp | 0
.../local-entities.cpp | 0
.../source-edit-generation/Inputs/empty-suite.json | 0
.../Inputs/two-function-suite.json | 0
.../source-edit-generation/Plugins/CMakeLists.txt | 0
.../Plugins/TestTransformationPlugin/CMakeLists.txt | 0
.../TestTransformationPlugin/TestTransformation.cpp | 0
.../source-edit-generation/Plugins/lit.local.cfg | 0
.../source-edit-generation/cli-errors.cpp | 0
.../source-edit-generation/coexistence.cpp | 0
.../cpp-bounded-buffers-replacements.cpp | 0
.../source-edit-generation/downgradable-errors.cpp | 0
.../source-edit-generation/happy-path.cpp | 0
.../source-edit-generation/lit.local.cfg | 0
.../multi-tu-conflict-edits-merge.cpp | 0
.../source-edit-generation/multi-tu-edits-merge.cpp | 0
.../source-edit-generation/write-failure.cpp | 0
.../ssaf-analyzer/Inputs/lu-badext.txt | 0
.../ssaf-analyzer/Inputs/lu-corrupt.json | 0
.../ssaf-analyzer/Inputs/lu-noext | 0
.../ssaf-analyzer/Inputs/lu-tags-only.json | 0
.../ssaf-analyzer/Inputs/lu-wrong-type.json | 0
.../ssaf-analyzer/Inputs/lu.json | 0
.../ssaf-analyzer/Outputs/all.json | 0
.../ssaf-analyzer/Outputs/both.json | 0
.../ssaf-analyzer/Outputs/empty-pairs.json | 0
.../ssaf-analyzer/Outputs/pairs.json | 0
.../ssaf-analyzer/Outputs/tags.json | 0
.../ssaf-analyzer/analyzer.test | 0
.../ssaf-analyzer/cli-errors.test | 0
.../ssaf-analyzer/help.test | 0
.../ssaf-analyzer/lit.local.cfg | 0
.../ssaf-analyzer/validation-errors-permissions.test | 0
.../ssaf-analyzer/validation-errors.test | 0
.../ssaf-analyzer/version.test | 0
.../ssaf-format/Artifact/Inputs/missing-type.json | 0
.../ssaf-format/Artifact/Inputs/rt-lu-empty.json | 0
.../Inputs/rt-multi-arch-shared-library-empty.json | 0
.../Inputs/rt-multi-arch-shared-library-nonempty.json | 0
.../Inputs/rt-multi-arch-static-library-empty.json | 0
.../Inputs/rt-multi-arch-static-library-nonempty.json | 0
.../Artifact/Inputs/rt-static-library-empty.json | 0
.../ssaf-format/Artifact/Inputs/rt-tu-empty.json | 0
.../ssaf-format/Artifact/Inputs/rt-wpa-empty.json | 0
.../ssaf-format/Artifact/Inputs/unknown-type.json | 0
.../ssaf-format/Artifact/round-trip.test | 0
.../ssaf-format/Artifact/top-level.test | 0
.../Inputs/CallGraph/invalid-direct-callee-element.json | 0
.../Inputs/CallGraph/invalid-direct-callee-id.json | 0
.../ssaf-format/Inputs/CallGraph/missing-def-col.json | 0
.../ssaf-format/Inputs/CallGraph/missing-def-file.json | 0
.../ssaf-format/Inputs/CallGraph/missing-def-line.json | 0
.../ssaf-format/Inputs/CallGraph/missing-def.json | 0
.../Inputs/CallGraph/missing-direct-callees.json | 0
.../Inputs/CallGraph/missing-pretty-name.json | 0
.../Inputs/CallGraph/missing-virtual-callees.json | 0
.../LUSummary/Inputs/data-element-not-object.json | 0
.../LUSummary/Inputs/data-entry-missing-data.json | 0
.../Inputs/data-entry-missing-summary-name.json | 0
.../ssaf-format/LUSummary/Inputs/data-not-array.json | 0
.../Inputs/duplicate-entity-id-in-data-map.json | 0
.../ssaf-format/LUSummary/Inputs/duplicate-entity.json | 0
.../LUSummary/Inputs/duplicate-summary-name.json | 0
.../LUSummary/Inputs/entity-data-element-not-object.json | 0
.../LUSummary/Inputs/entity-data-missing-entity-id.json | 0
.../Inputs/entity-data-missing-entity-summary.json | 0
.../LUSummary/Inputs/entity-id-not-uint64.json | 0
.../LUSummary/Inputs/entity-name-missing-namespace.json | 0
.../LUSummary/Inputs/entity-name-missing-suffix.json | 0
.../LUSummary/Inputs/entity-name-missing-usr.json | 0
.../LUSummary/Inputs/id-table-element-not-object.json | 0
.../LUSummary/Inputs/id-table-entry-id-not-uint64.json | 0
.../LUSummary/Inputs/id-table-entry-missing-id.json | 0
.../LUSummary/Inputs/id-table-entry-missing-name.json | 0
.../ssaf-format/LUSummary/Inputs/id-table-not-array.json | 0
.../ssaf-format/LUSummary/Inputs/invalid-syntax.json | 0
.../LUSummary/Inputs/linkage-table-duplicate-id.json | 0
.../Inputs/linkage-table-element-not-object.json | 0
.../Inputs/linkage-table-entry-id-not-uint64.json | 0
.../Inputs/linkage-table-entry-linkage-invalid-type.json | 0
.../Inputs/linkage-table-entry-linkage-missing-type.json | 0
.../LUSummary/Inputs/linkage-table-entry-missing-id.json | 0
.../Inputs/linkage-table-entry-missing-linkage.json | 0
.../LUSummary/Inputs/linkage-table-extra-id.json | 0
.../LUSummary/Inputs/linkage-table-missing-id.json | 0
.../LUSummary/Inputs/linkage-table-not-array.json | 0
.../Inputs/lu-namespace-element-invalid-kind.json | 0
.../Inputs/lu-namespace-element-missing-kind.json | 0
.../Inputs/lu-namespace-element-missing-name.json | 0
.../Inputs/lu-namespace-element-not-object.json | 0
.../LUSummary/Inputs/lu-namespace-not-array.json | 0
.../ssaf-format/LUSummary/Inputs/mismatched-type.json | 0
.../ssaf-format/LUSummary/Inputs/missing-data.json | 0
.../ssaf-format/LUSummary/Inputs/missing-id-table.json | 0
.../LUSummary/Inputs/missing-linkage-table.json | 0
.../LUSummary/Inputs/missing-lu-namespace.json | 0
.../LUSummary/Inputs/missing-target-triple.json | 0
.../ssaf-format/LUSummary/Inputs/missing-type.json | 0
.../LUSummary/Inputs/namespace-element-invalid-kind.json | 0
.../LUSummary/Inputs/namespace-element-missing-kind.json | 0
.../LUSummary/Inputs/namespace-element-missing-name.json | 0
.../LUSummary/Inputs/namespace-element-not-object.json | 0
.../ssaf-format/LUSummary/Inputs/not-json-extension.txt | 0
.../LUSummary/Inputs/not-normalized-target-triple.json | 0
.../ssaf-format/LUSummary/Inputs/not-object.json | 0
.../LUSummary/Inputs/pairs-element-not-object.json | 0
.../LUSummary/Inputs/pairs-invalid-first-field.json | 0
.../LUSummary/Inputs/pairs-invalid-pairs-field-type.json | 0
.../LUSummary/Inputs/pairs-invalid-second-field.json | 0
.../LUSummary/Inputs/pairs-missing-first-field.json | 0
.../LUSummary/Inputs/pairs-missing-pairs-field.json | 0
.../LUSummary/Inputs/pairs-missing-second-field.json | 0
.../Inputs/read-entity-summary-no-format-info.json | 0
.../LUSummary/Inputs/rt-empty-data-entry.json | 0
.../ssaf-format/LUSummary/Inputs/rt-empty-namespace.json | 0
.../LUSummary/Inputs/rt-linkage-external.json | 0
.../LUSummary/Inputs/rt-linkage-internal.json | 0
.../LUSummary/Inputs/rt-linkage-multiple.json | 0
.../ssaf-format/LUSummary/Inputs/rt-linkage-none.json | 0
.../LUSummary/Inputs/rt-multiple-namespace-elements.json | 0
.../LUSummary/Inputs/rt-single-namespace-element.json | 0
.../LUSummary/Inputs/rt-two-summary-types.json | 0
.../ssaf-format/LUSummary/id-table.test | 0
.../ssaf-format/LUSummary/io.test | 0
.../ssaf-format/LUSummary/linkage.test | 0
.../ssaf-format/LUSummary/permissions.test | 0
.../ssaf-format/LUSummary/round-trip.test | 0
.../ssaf-format/LUSummary/summary-data.test | 0
.../ssaf-format/LUSummary/top-level.test | 0
.../Inputs/duplicate-target-triple.json | 0
.../MultiArchSharedLibrary/Inputs/invalid-syntax.json | 0
.../Inputs/member-empty-lu-namespace.json | 0
.../Inputs/member-inner-error.json | 0
.../Inputs/member-mismatched-type.json | 0
.../Inputs/member-missing-type.json | 0
.../Inputs/member-namespace-mismatch.json | 0
.../MultiArchSharedLibrary/Inputs/member-not-object.json | 0
.../MultiArchSharedLibrary/Inputs/members-not-array.json | 0
.../MultiArchSharedLibrary/Inputs/mismatched-type.json | 0
.../MultiArchSharedLibrary/Inputs/missing-members.json | 0
.../MultiArchSharedLibrary/Inputs/missing-namespace.json | 0
.../MultiArchSharedLibrary/Inputs/missing-type.json | 0
.../Inputs/namespace-not-array.json | 0
.../MultiArchSharedLibrary/Inputs/not-json-extension.txt | 0
.../MultiArchSharedLibrary/Inputs/not-object.json | 0
.../MultiArchSharedLibrary/Inputs/rt-empty-members.json | 0
.../Inputs/rt-multiple-members.json | 0
.../Inputs/rt-nonempty-members.json | 0
.../MultiArchSharedLibrary/Inputs/rt-single-member.json | 0
.../MultiArchSharedLibrary/Inputs/rt-three-members.json | 0
.../Inputs/unsorted-members-input.json | 0
.../Inputs/unsorted-three-members-input.json | 0
.../ssaf-format/MultiArchSharedLibrary/io.test | 0
.../ssaf-format/MultiArchSharedLibrary/permissions.test | 0
.../ssaf-format/MultiArchSharedLibrary/round-trip.test | 0
.../ssaf-format/MultiArchSharedLibrary/top-level.test | 0
.../Inputs/duplicate-target-triple.json | 0
.../MultiArchStaticLibrary/Inputs/invalid-syntax.json | 0
.../Inputs/member-inner-error.json | 0
.../Inputs/member-mismatched-type.json | 0
.../Inputs/member-missing-type.json | 0
.../Inputs/member-namespace-mismatch.json | 0
.../MultiArchStaticLibrary/Inputs/member-not-object.json | 0
.../MultiArchStaticLibrary/Inputs/members-not-array.json | 0
.../MultiArchStaticLibrary/Inputs/mismatched-type.json | 0
.../MultiArchStaticLibrary/Inputs/missing-members.json | 0
.../MultiArchStaticLibrary/Inputs/missing-namespace.json | 0
.../MultiArchStaticLibrary/Inputs/missing-type.json | 0
.../Inputs/namespace-not-object.json | 0
.../Inputs/namespace-wrong-kind.json | 0
.../MultiArchStaticLibrary/Inputs/not-json-extension.txt | 0
.../MultiArchStaticLibrary/Inputs/not-object.json | 0
.../MultiArchStaticLibrary/Inputs/rt-empty-members.json | 0
.../Inputs/rt-multiple-members.json | 0
.../Inputs/rt-nonempty-members.json | 0
.../MultiArchStaticLibrary/Inputs/rt-single-member.json | 0
.../MultiArchStaticLibrary/Inputs/rt-three-members.json | 0
.../Inputs/unsorted-members-input.json | 0
.../Inputs/unsorted-three-members-input.json | 0
.../ssaf-format/MultiArchStaticLibrary/io.test | 0
.../ssaf-format/MultiArchStaticLibrary/permissions.test | 0
.../ssaf-format/MultiArchStaticLibrary/round-trip.test | 0
.../ssaf-format/MultiArchStaticLibrary/top-level.test | 0
.../StaticLibrary/Inputs/duplicate-member.json | 0
.../ssaf-format/StaticLibrary/Inputs/invalid-syntax.json | 0
.../StaticLibrary/Inputs/member-inner-error.json | 0
.../StaticLibrary/Inputs/member-mismatched-type.json | 0
.../StaticLibrary/Inputs/member-missing-type.json | 0
.../StaticLibrary/Inputs/member-not-object.json | 0
.../StaticLibrary/Inputs/member-triple-mismatch.json | 0
.../StaticLibrary/Inputs/members-not-array.json | 0
.../StaticLibrary/Inputs/mismatched-type.json | 0
.../StaticLibrary/Inputs/missing-members.json | 0
.../StaticLibrary/Inputs/missing-namespace.json | 0
.../StaticLibrary/Inputs/missing-target-triple.json | 0
.../ssaf-format/StaticLibrary/Inputs/missing-type.json | 0
.../StaticLibrary/Inputs/namespace-invalid-kind.json | 0
.../StaticLibrary/Inputs/namespace-missing-kind.json | 0
.../StaticLibrary/Inputs/namespace-missing-name.json | 0
.../StaticLibrary/Inputs/namespace-wrong-kind.json | 0
.../StaticLibrary/Inputs/not-json-extension.txt | 0
.../Inputs/not-normalized-target-triple.json | 0
.../ssaf-format/StaticLibrary/Inputs/not-object.json | 0
.../ssaf-format/StaticLibrary/Inputs/rt-empty.json | 0
.../StaticLibrary/Inputs/rt-multiple-members.json | 0
.../StaticLibrary/Inputs/rt-single-member.json | 0
.../StaticLibrary/Inputs/unsorted-members-input.json | 0
.../ssaf-format/StaticLibrary/io.test | 0
.../ssaf-format/StaticLibrary/permissions.test | 0
.../ssaf-format/StaticLibrary/round-trip.test | 0
.../ssaf-format/StaticLibrary/top-level.test | 0
.../TUSummary/Inputs/data-element-not-object.json | 0
.../TUSummary/Inputs/data-entry-missing-data.json | 0
.../Inputs/data-entry-missing-summary-name.json | 0
.../ssaf-format/TUSummary/Inputs/data-not-array.json | 0
.../Inputs/duplicate-entity-id-in-data-map.json | 0
.../ssaf-format/TUSummary/Inputs/duplicate-entity.json | 0
.../TUSummary/Inputs/duplicate-summary-name.json | 0
.../TUSummary/Inputs/entity-data-element-not-object.json | 0
.../TUSummary/Inputs/entity-data-missing-entity-id.json | 0
.../Inputs/entity-data-missing-entity-summary.json | 0
.../TUSummary/Inputs/entity-id-not-uint64.json | 0
.../TUSummary/Inputs/entity-name-missing-suffix.json | 0
.../TUSummary/Inputs/entity-name-missing-usr.json | 0
.../TUSummary/Inputs/id-table-element-not-object.json | 0
.../TUSummary/Inputs/id-table-entry-id-not-uint64.json | 0
.../TUSummary/Inputs/id-table-entry-missing-id.json | 0
.../TUSummary/Inputs/id-table-entry-missing-name.json | 0
.../ssaf-format/TUSummary/Inputs/id-table-not-array.json | 0
.../ssaf-format/TUSummary/Inputs/invalid-kind.json | 0
.../ssaf-format/TUSummary/Inputs/invalid-syntax.json | 0
.../TUSummary/Inputs/linkage-table-duplicate-id.json | 0
.../Inputs/linkage-table-element-not-object.json | 0
.../Inputs/linkage-table-entry-id-not-uint64.json | 0
.../Inputs/linkage-table-entry-linkage-invalid-type.json | 0
.../Inputs/linkage-table-entry-linkage-missing-type.json | 0
.../TUSummary/Inputs/linkage-table-entry-missing-id.json | 0
.../Inputs/linkage-table-entry-missing-linkage.json | 0
.../TUSummary/Inputs/linkage-table-extra-id.json | 0
.../TUSummary/Inputs/linkage-table-missing-id.json | 0
.../TUSummary/Inputs/linkage-table-not-array.json | 0
.../ssaf-format/TUSummary/Inputs/mismatched-type.json | 0
.../ssaf-format/TUSummary/Inputs/missing-data.json | 0
.../ssaf-format/TUSummary/Inputs/missing-id-table.json | 0
.../ssaf-format/TUSummary/Inputs/missing-kind.json | 0
.../TUSummary/Inputs/missing-linkage-table.json | 0
.../ssaf-format/TUSummary/Inputs/missing-name.json | 0
.../TUSummary/Inputs/missing-target-triple.json | 0
.../TUSummary/Inputs/missing-tu-namespace.json | 0
.../ssaf-format/TUSummary/Inputs/missing-type.json | 0
.../ssaf-format/TUSummary/Inputs/not-json-extension.txt | 0
.../TUSummary/Inputs/not-normalized-target-triple.json | 0
.../ssaf-format/TUSummary/Inputs/not-object.json | 0
.../TUSummary/Inputs/pairs-element-not-object.json | 0
.../TUSummary/Inputs/pairs-invalid-first-field.json | 0
.../TUSummary/Inputs/pairs-invalid-pairs-field-type.json | 0
.../TUSummary/Inputs/pairs-invalid-second-field.json | 0
.../TUSummary/Inputs/pairs-missing-first-field.json | 0
.../TUSummary/Inputs/pairs-missing-pairs-field.json | 0
.../TUSummary/Inputs/pairs-missing-second-field.json | 0
.../Inputs/read-entity-summary-no-format-info.json | 0
.../TUSummary/Inputs/rt-empty-data-entry.json | 0
.../ssaf-format/TUSummary/Inputs/rt-empty.json | 0
.../ssaf-format/TUSummary/Inputs/rt-link-unit.json | 0
.../TUSummary/Inputs/rt-linkage-external.json | 0
.../TUSummary/Inputs/rt-linkage-internal.json | 0
.../TUSummary/Inputs/rt-linkage-multiple.json | 0
.../ssaf-format/TUSummary/Inputs/rt-linkage-none.json | 0
.../TUSummary/Inputs/rt-two-summary-types.json | 0
.../ssaf-format/TUSummary/id-table.test | 0
.../ssaf-format/TUSummary/io.test | 0
.../ssaf-format/TUSummary/linkage.test | 0
.../ssaf-format/TUSummary/permissions.test | 0
.../ssaf-format/TUSummary/round-trip.test | 0
.../ssaf-format/TUSummary/summary-data.test | 0
.../ssaf-format/TUSummary/top-level.test | 0
.../WPASuite/Inputs/duplicate-analysis-name.json | 0
.../ssaf-format/WPASuite/Inputs/mismatched-type.json | 0
.../ssaf-format/WPASuite/Inputs/missing-results.json | 0
.../ssaf-format/WPASuite/Inputs/missing-type.json | 0
.../WPASuite/Inputs/pairs-result-bad-count.json | 0
.../WPASuite/Inputs/pairs-result-bad-pair-counts.json | 0
.../WPASuite/Inputs/pairs-result-entry-not-object.json | 0
.../WPASuite/Inputs/pairs-result-missing-entity-id.json | 0
.../Inputs/result-entry-analysis-name-not-string.json | 0
.../WPASuite/Inputs/result-entry-deserializer-error.json | 0
.../Inputs/result-entry-missing-analysis-name.json | 0
.../Inputs/result-entry-missing-result-field.json | 0
.../WPASuite/Inputs/result-entry-no-format-info.json | 0
.../WPASuite/Inputs/result-entry-not-object.json | 0
.../WPASuite/Inputs/result-entry-result-not-object.json | 0
.../ssaf-format/WPASuite/Inputs/results-not-array.json | 0
.../ssaf-format/WPASuite/Inputs/rt-all-results.json | 0
.../ssaf-format/WPASuite/Inputs/rt-empty.json | 0
.../ssaf-format/WPASuite/Inputs/rt-multiple-results.json | 0
.../WPASuite/Inputs/tags-result-bad-element.json | 0
.../WPASuite/Inputs/tags-result-missing-tags.json | 0
.../WPASuite/Inputs/tagspairs-result-bad-field-type.json | 0
.../WPASuite/Inputs/tagspairs-result-missing-field.json | 0
.../ssaf-format/WPASuite/with-plugin.test | 0
.../ssaf-format/WPASuite/without-plugin.test | 0
.../ssaf-format/call-graph-invalid-json-format.test | 0
.../ssaf-format/list.test | 0
.../ssaf-format/lit.local.cfg | 0
.../ssaf-linker/Inputs/bad-artifact.json | 0
.../ssaf-linker/Inputs/lib-aarch64.json | 0
.../ssaf-linker/Inputs/lib-arm64.json | 0
.../ssaf-linker/Inputs/lib-ios-versioned.json | 0
.../ssaf-linker/Inputs/lib-ios.json | 0
.../ssaf-linker/Inputs/lib-otherns.json | 0
.../ssaf-linker/Inputs/lib-x86_64.json | 0
.../ssaf-linker/Inputs/libfoo-2arch.json | 0
.../ssaf-linker/Inputs/libmulti-empty-wrapper.json | 0
.../ssaf-linker/Inputs/libmulti-multiarch-wrongns.json | 0
.../ssaf-linker/Inputs/libord-reversed.json | 0
.../ssaf-linker/Inputs/libshared-empty-wrapper.json | 0
.../ssaf-linker/Inputs/libshared-multiarch-wrongns.json | 0
.../ssaf-linker/Inputs/libtwo-2arch.json | 0
.../ssaf-linker/Inputs/lu-arm64.json | 0
.../ssaf-linker/Inputs/lu-ios-libfoo.json | 0
.../ssaf-linker/Inputs/lu-macosx-libfoo.json | 0
.../ssaf-linker/Inputs/lu-otherns.json | 0
.../ssaf-linker/Inputs/lu-x86_64.json | 0
.../ssaf-linker/Inputs/tu-1.json | 0
.../ssaf-linker/Inputs/tu-2.json | 0
.../ssaf-linker/Inputs/tu-badext.txt | 0
.../ssaf-linker/Inputs/tu-empty.json | 0
.../Inputs/tu-invalid-entity-id-multikey.json | 0
.../ssaf-linker/Inputs/tu-invalid-entity-id-ref.json | 0
.../ssaf-linker/Inputs/tu-invalid-entity-id-value.json | 0
.../ssaf-linker/Inputs/tu-linux.json | 0
.../ssaf-linker/Inputs/tu-malformed.json | 0
.../ssaf-linker/Inputs/tu-missing-fields.json | 0
.../ssaf-linker/Inputs/tu-noext | 0
.../ssaf-linker/Inputs/tu-x86_64.json | 0
.../ssaf-linker/Outputs/lib-explicit-triple.json | 0
.../ssaf-linker/Outputs/lib-namespace-override.json | 0
.../ssaf-linker/Outputs/libempty.json | 0
.../ssaf-linker/Outputs/libfoo-3arch.json | 0
.../ssaf-linker/Outputs/libmulti-1arch.json | 0
.../ssaf-linker/Outputs/libmulti-2arch.json | 0
.../ssaf-linker/Outputs/libmulti-3arch.json | 0
.../ssaf-linker/Outputs/libshared-1arch.json | 0
.../ssaf-linker/Outputs/libshared-2arch.json | 0
.../ssaf-linker/Outputs/libtwo.json | 0
.../ssaf-linker/Outputs/lu-1+2.json | 0
.../ssaf-linker/Outputs/lu-1.json | 0
.../ssaf-linker/Outputs/lu-2.json | 0
.../ssaf-linker/Outputs/lu-empty.json | 0
.../ssaf-linker/cli.test | 0
.../ssaf-linker/help.test | 0
.../ssaf-linker/io.test | 0
.../ssaf-linker/linking-errors.test | 0
.../ssaf-linker/linking.test | 0
.../ssaf-linker/multi-arch-create.test | 0
.../ssaf-linker/multi-arch-help.test | 0
.../ssaf-linker/multi-arch-time.test | 0
.../ssaf-linker/multi-arch-verbose.test | 0
.../ssaf-linker/static-library-create.test | 0
.../ssaf-linker/static-library-help.test | 0
.../ssaf-linker/static-library-time.test | 0
.../ssaf-linker/static-library-verbose.test | 0
.../ssaf-linker/time.test | 0
.../ssaf-linker/validation-errors-permissions.test | 0
.../ssaf-linker/validation-errors.test | 0
.../ssaf-linker/verbose.test | 0
.../ssaf-linker/version.test | 0
.../Inputs/conflict-text-q-offset-30.tpl.yaml | 0
.../ssaf-src-edit-merge/Inputs/conflict-text-y.tpl.yaml | 0
.../ssaf-src-edit-merge/Inputs/conflict-text-z.tpl.yaml | 0
.../ssaf-src-edit-merge/Inputs/empty.tpl.yaml | 0
.../Inputs/internal-conflict.tpl.yaml | 0
.../Inputs/non-overlapping-a.tpl.yaml | 0
.../Inputs/non-overlapping-b.tpl.yaml | 0
.../Inputs/range-overlap-cluster-of-three.tpl.yaml | 0
.../Inputs/range-overlap-distinct-offsets.tpl.yaml | 0
.../Inputs/single-non-zero-length-replacement.tpl.yaml | 0
.../Inputs/single-replacement-offset-30.tpl.yaml | 0
.../Inputs/single-replacement.tpl.yaml | 0
.../Inputs/single-zero-length-insert.tpl.yaml | 0
.../ssaf-src-edit-merge/bad-args.test | 0
.../conflict-preserves-unrelated-zero-length-insert.test | 0
.../conflict-stderr-without-sarif.test | 0
.../ssaf-src-edit-merge/empty-input.test | 0
.../ssaf-src-edit-merge/help.test | 0
.../ssaf-src-edit-merge/identical-overlap.test | 0
.../ssaf-src-edit-merge/malformed-yaml.test | 0
.../ssaf-src-edit-merge/no-conflict-sarif-empty.test | 0
.../ssaf-src-edit-merge/non-overlapping.test | 0
.../ssaf-src-edit-merge/order-invariance.test | 0
.../ssaf-src-edit-merge/parent-dir-missing.test | 0
.../pre-dedup-identical-non-zero-length.test | 0
.../pre-dedup-zero-length-insert.test | 0
.../range-overlap-cluster-of-three-sarif.test | 0
.../range-overlap-conflict-sarif.test | 0
.../sarif-out-parent-dir-missing.test | 0
.../single-input-internal-conflict.test | 0
.../ssaf-src-edit-merge/single-input.test | 0
.../ssaf-src-edit-merge/three-input-all-conflict.test | 0
.../ssaf-src-edit-merge/three-input-conflict-sarif.test | 0
.../two-conflicts-different-offsets-sarif.test | 0
.../ssaf-src-edit-merge/two-input-conflict-sarif.test | 0
.../ssaf-src-edit-merge/two-input-conflict.test | 0
clang/unittests/ScalableStaticAnalysis/CMakeLists.txt | 9 +++++++++
454 files changed, 11 insertions(+), 2 deletions(-)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/tu-summary-bad-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/tu-summary-bad-summary.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/tu-summary-no-key.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/tu-summary.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/wpa-result-bad-edges.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/wpa-result-bad-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/wpa-result-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/wpa-result-no-key.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/wpa-result-odd-count.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/Inputs/wpa-result.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/entity-name-no-conflict.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/external-inline-function-in-multi-tu.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/lref-to-rref-cast.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/multi-decl-contributor.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/multi-dim-pointer-flow-constraint.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/system-header-opt-out.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/tu-summary-serialization.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/unsafe-buffer-reachable-cast-cycle.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/PointerFlow/wpa-result-serialization.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/Inputs/tu-summary-bad-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/Inputs/tu-summary-no-key.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/Inputs/tu-summary.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/Inputs/wpa-result-bad-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/Inputs/wpa-result-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/Inputs/wpa-result-no-key.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/tu-summary-serialization.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/type-constrained-pointers.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-main.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-new-delete.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/TypeConstrainedPointers/wpa-result-serialization.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/tu-summary-bad-element.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/tu-summary-bad-ptr-level.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/tu-summary-no-key.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/tu-summary.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/wpa-result-bad-epls.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/wpa-result-bad-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/wpa-result-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/wpa-result-no-key.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/wpa-result-odd-count.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/Inputs/wpa-result.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/tu-summary-serialization.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/UnsafeBufferUsage/wpa-result-serialization.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/call-graph-local-entities.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/call-graph.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/cli-errors-compilation-unit-id.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/command-line-interface.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/downgradable-errors.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/extraction-works-alongside-compilation.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/help.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/local-entities.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/Inputs/empty-suite.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/Inputs/two-function-suite.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/Plugins/CMakeLists.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/Plugins/TestTransformationPlugin/CMakeLists.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/Plugins/TestTransformationPlugin/TestTransformation.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/Plugins/lit.local.cfg (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/cli-errors.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/coexistence.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/cpp-bounded-buffers-replacements.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/downgradable-errors.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/happy-path.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/lit.local.cfg (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/multi-tu-conflict-edits-merge.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/multi-tu-edits-merge.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/source-edit-generation/write-failure.cpp (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Inputs/lu-badext.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Inputs/lu-corrupt.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Inputs/lu-noext (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Inputs/lu-tags-only.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Inputs/lu-wrong-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Inputs/lu.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Outputs/all.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Outputs/both.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Outputs/empty-pairs.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Outputs/pairs.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/Outputs/tags.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/analyzer.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/cli-errors.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/help.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/lit.local.cfg (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/validation-errors-permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/validation-errors.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-analyzer/version.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-lu-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-nonempty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-nonempty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-static-library-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-tu-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/rt-wpa-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/Inputs/unknown-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/round-trip.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Artifact/top-level.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/invalid-direct-callee-element.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/invalid-direct-callee-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-def-col.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-def-file.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-def-line.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-def.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-direct-callees.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-pretty-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/Inputs/CallGraph/missing-virtual-callees.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/data-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/data-entry-missing-data.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/data-entry-missing-summary-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/data-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/duplicate-entity-id-in-data-map.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/duplicate-entity.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/duplicate-summary-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-data-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-summary.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-id-not-uint64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-name-missing-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-name-missing-suffix.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/entity-name-missing-usr.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/id-table-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/id-table-entry-id-not-uint64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/id-table-entry-missing-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/id-table-entry-missing-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/id-table-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/invalid-syntax.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-duplicate-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-entry-id-not-uint64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-linkage.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-extra-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-missing-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/linkage-table-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/lu-namespace-element-invalid-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/lu-namespace-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/lu-namespace-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/missing-data.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/missing-id-table.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/missing-linkage-table.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/missing-lu-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/missing-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/namespace-element-invalid-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/namespace-element-missing-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/namespace-element-missing-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/namespace-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/not-json-extension.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/not-normalized-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-invalid-first-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-invalid-pairs-field-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-invalid-second-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-missing-first-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-missing-pairs-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/pairs-missing-second-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/read-entity-summary-no-format-info.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-empty-data-entry.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-empty-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-linkage-external.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-linkage-internal.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-linkage-multiple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-linkage-none.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-multiple-namespace-elements.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-single-namespace-element.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/Inputs/rt-two-summary-types.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/id-table.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/io.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/linkage.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/round-trip.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/summary-data.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/LUSummary/top-level.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/duplicate-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/invalid-syntax.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/member-empty-lu-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/member-inner-error.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/member-mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/member-missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/member-namespace-mismatch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/member-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/members-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/missing-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/missing-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/namespace-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/not-json-extension.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/rt-empty-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/rt-multiple-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/rt-nonempty-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/rt-single-member.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/rt-three-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-members-input.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-three-members-input.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/io.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/round-trip.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchSharedLibrary/top-level.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/duplicate-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/invalid-syntax.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/member-inner-error.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/member-mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/member-missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/member-namespace-mismatch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/member-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/members-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/missing-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/missing-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-wrong-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/not-json-extension.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/rt-empty-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/rt-multiple-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/rt-nonempty-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/rt-single-member.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/rt-three-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-members-input.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-three-members-input.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/io.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/round-trip.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/MultiArchStaticLibrary/top-level.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/duplicate-member.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/invalid-syntax.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/member-inner-error.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/member-mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/member-missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/member-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/member-triple-mismatch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/members-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/missing-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/missing-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/missing-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/namespace-invalid-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/namespace-missing-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/namespace-missing-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/namespace-wrong-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/not-json-extension.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/not-normalized-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/rt-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/rt-multiple-members.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/rt-single-member.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/Inputs/unsorted-members-input.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/io.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/round-trip.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/StaticLibrary/top-level.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/data-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/data-entry-missing-data.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/data-entry-missing-summary-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/data-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/duplicate-entity-id-in-data-map.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/duplicate-entity.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/duplicate-summary-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/entity-data-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-summary.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/entity-id-not-uint64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/entity-name-missing-suffix.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/entity-name-missing-usr.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/id-table-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/id-table-entry-id-not-uint64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/id-table-entry-missing-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/id-table-entry-missing-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/id-table-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/invalid-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/invalid-syntax.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-duplicate-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-entry-id-not-uint64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-linkage.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-extra-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-missing-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/linkage-table-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-data.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-id-table.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-kind.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-linkage-table.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-tu-namespace.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/not-json-extension.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/not-normalized-target-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-element-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-invalid-first-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-invalid-pairs-field-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-invalid-second-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-missing-first-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-missing-pairs-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/pairs-missing-second-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/read-entity-summary-no-format-info.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-empty-data-entry.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-link-unit.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-linkage-external.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-linkage-internal.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-linkage-multiple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-linkage-none.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/Inputs/rt-two-summary-types.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/id-table.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/io.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/linkage.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/round-trip.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/summary-data.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/TUSummary/top-level.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/duplicate-analysis-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/mismatched-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/missing-results.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/missing-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/pairs-result-bad-count.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/pairs-result-bad-pair-counts.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/pairs-result-entry-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/pairs-result-missing-entity-id.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-analysis-name-not-string.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-deserializer-error.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-missing-analysis-name.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-missing-result-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-no-format-info.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/result-entry-result-not-object.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/results-not-array.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/rt-all-results.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/rt-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/rt-multiple-results.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/tags-result-bad-element.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/tags-result-missing-tags.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/tagspairs-result-bad-field-type.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/Inputs/tagspairs-result-missing-field.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/with-plugin.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/WPASuite/without-plugin.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/call-graph-invalid-json-format.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/list.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-format/lit.local.cfg (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/bad-artifact.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lib-aarch64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lib-arm64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lib-ios-versioned.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lib-ios.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lib-otherns.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lib-x86_64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libfoo-2arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libmulti-empty-wrapper.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libmulti-multiarch-wrongns.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libord-reversed.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libshared-empty-wrapper.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libshared-multiarch-wrongns.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/libtwo-2arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lu-arm64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lu-ios-libfoo.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lu-macosx-libfoo.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lu-otherns.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/lu-x86_64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-1.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-2.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-badext.txt (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-invalid-entity-id-multikey.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-invalid-entity-id-ref.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-invalid-entity-id-value.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-linux.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-malformed.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-missing-fields.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-noext (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Inputs/tu-x86_64.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/lib-explicit-triple.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/lib-namespace-override.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libempty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libfoo-3arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libmulti-1arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libmulti-2arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libmulti-3arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libshared-1arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libshared-2arch.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/libtwo.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/lu-1+2.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/lu-1.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/lu-2.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/Outputs/lu-empty.json (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/cli.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/help.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/io.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/linking-errors.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/linking.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/multi-arch-create.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/multi-arch-help.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/multi-arch-time.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/multi-arch-verbose.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/static-library-create.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/static-library-help.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/static-library-time.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/static-library-verbose.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/time.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/validation-errors-permissions.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/validation-errors.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/verbose.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-linker/version.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/conflict-text-q-offset-30.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/conflict-text-y.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/conflict-text-z.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/empty.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/internal-conflict.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/non-overlapping-a.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/non-overlapping-b.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/range-overlap-cluster-of-three.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/range-overlap-distinct-offsets.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/single-non-zero-length-replacement.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/single-replacement-offset-30.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/single-replacement.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/Inputs/single-zero-length-insert.tpl.yaml (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/bad-args.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/conflict-preserves-unrelated-zero-length-insert.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/conflict-stderr-without-sarif.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/empty-input.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/help.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/identical-overlap.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/malformed-yaml.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/no-conflict-sarif-empty.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/non-overlapping.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/order-invariance.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/parent-dir-missing.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/pre-dedup-identical-non-zero-length.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/pre-dedup-zero-length-insert.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/range-overlap-cluster-of-three-sarif.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/range-overlap-conflict-sarif.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/sarif-out-parent-dir-missing.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/single-input-internal-conflict.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/single-input.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/three-input-all-conflict.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/three-input-conflict-sarif.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/two-conflicts-different-offsets-sarif.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/two-input-conflict-sarif.test (100%)
rename clang/test/{Analysis/Scalable => ScalableStaticAnalysis}/ssaf-src-edit-merge/two-input-conflict.test (100%)
diff --git a/.github/new-prs-labeler.yml b/.github/new-prs-labeler.yml
index 4269b76271b942..c818db8556d5d0 100644
--- a/.github/new-prs-labeler.yml
+++ b/.github/new-prs-labeler.yml
@@ -708,7 +708,7 @@ clang:ssaf:
- clang/docs/ScalableStaticAnalysis/**
- clang/include/clang/ScalableStaticAnalysis/**
- clang/lib/ScalableStaticAnalysis/**
- - clang/test/Analysis/Scalable/**
+ - clang/test/ScalableStaticAnalysis/**
- clang/unittests/ScalableStaticAnalysis/**
clang:static analyzer:
diff --git a/clang/test/CMakeLists.txt b/clang/test/CMakeLists.txt
index 7fff733d8ee7b2..7dbc35baf9e7b4 100644
--- a/clang/test/CMakeLists.txt
+++ b/clang/test/CMakeLists.txt
@@ -221,7 +221,7 @@ if(CLANG_PLUGIN_SUPPORT AND LLVM_ENABLE_PLUGINS AND NOT WIN32)
SSAFExamplePlugin
SSAFTestTransformationPlugin
)
- add_subdirectory(Analysis/Scalable/source-edit-generation/Plugins)
+ add_subdirectory(ScalableStaticAnalysis/source-edit-generation/Plugins)
endif()
if (HAVE_CLANG_REPL_SUPPORT)
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary-bad-array.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary-bad-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary-bad-array.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary-bad-array.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary-bad-summary.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary-bad-summary.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary-bad-summary.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary-bad-summary.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary-no-key.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary-no-key.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary-no-key.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary-no-key.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/tu-summary.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/tu-summary.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-bad-edges.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-bad-edges.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-bad-edges.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-bad-edges.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-bad-id.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-bad-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-bad-id.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-bad-id.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-empty.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-empty.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-empty.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-no-key.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-no-key.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-no-key.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-no-key.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-odd-count.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-odd-count.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result-odd-count.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result-odd-count.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result.json b/clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result.json
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/Inputs/wpa-result.json
rename to clang/test/ScalableStaticAnalysis/PointerFlow/Inputs/wpa-result.json
diff --git a/clang/test/Analysis/Scalable/PointerFlow/entity-name-no-conflict.cpp b/clang/test/ScalableStaticAnalysis/PointerFlow/entity-name-no-conflict.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/entity-name-no-conflict.cpp
rename to clang/test/ScalableStaticAnalysis/PointerFlow/entity-name-no-conflict.cpp
diff --git a/clang/test/Analysis/Scalable/PointerFlow/external-inline-function-in-multi-tu.test b/clang/test/ScalableStaticAnalysis/PointerFlow/external-inline-function-in-multi-tu.test
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/external-inline-function-in-multi-tu.test
rename to clang/test/ScalableStaticAnalysis/PointerFlow/external-inline-function-in-multi-tu.test
diff --git a/clang/test/Analysis/Scalable/PointerFlow/lref-to-rref-cast.test b/clang/test/ScalableStaticAnalysis/PointerFlow/lref-to-rref-cast.test
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/lref-to-rref-cast.test
rename to clang/test/ScalableStaticAnalysis/PointerFlow/lref-to-rref-cast.test
diff --git a/clang/test/Analysis/Scalable/PointerFlow/multi-decl-contributor.cpp b/clang/test/ScalableStaticAnalysis/PointerFlow/multi-decl-contributor.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/multi-decl-contributor.cpp
rename to clang/test/ScalableStaticAnalysis/PointerFlow/multi-decl-contributor.cpp
diff --git a/clang/test/Analysis/Scalable/PointerFlow/multi-dim-pointer-flow-constraint.test b/clang/test/ScalableStaticAnalysis/PointerFlow/multi-dim-pointer-flow-constraint.test
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/multi-dim-pointer-flow-constraint.test
rename to clang/test/ScalableStaticAnalysis/PointerFlow/multi-dim-pointer-flow-constraint.test
diff --git a/clang/test/Analysis/Scalable/PointerFlow/system-header-opt-out.cpp b/clang/test/ScalableStaticAnalysis/PointerFlow/system-header-opt-out.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/system-header-opt-out.cpp
rename to clang/test/ScalableStaticAnalysis/PointerFlow/system-header-opt-out.cpp
diff --git a/clang/test/Analysis/Scalable/PointerFlow/tu-summary-serialization.test b/clang/test/ScalableStaticAnalysis/PointerFlow/tu-summary-serialization.test
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/tu-summary-serialization.test
rename to clang/test/ScalableStaticAnalysis/PointerFlow/tu-summary-serialization.test
diff --git a/clang/test/Analysis/Scalable/PointerFlow/unsafe-buffer-reachable-cast-cycle.test b/clang/test/ScalableStaticAnalysis/PointerFlow/unsafe-buffer-reachable-cast-cycle.test
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/unsafe-buffer-reachable-cast-cycle.test
rename to clang/test/ScalableStaticAnalysis/PointerFlow/unsafe-buffer-reachable-cast-cycle.test
diff --git a/clang/test/Analysis/Scalable/PointerFlow/wpa-result-serialization.test b/clang/test/ScalableStaticAnalysis/PointerFlow/wpa-result-serialization.test
similarity index 100%
rename from clang/test/Analysis/Scalable/PointerFlow/wpa-result-serialization.test
rename to clang/test/ScalableStaticAnalysis/PointerFlow/wpa-result-serialization.test
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/tu-summary-bad-id.json b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/tu-summary-bad-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/tu-summary-bad-id.json
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/tu-summary-bad-id.json
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/tu-summary-no-key.json b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/tu-summary-no-key.json
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/tu-summary-no-key.json
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/tu-summary-no-key.json
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/tu-summary.json b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/tu-summary.json
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/tu-summary.json
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/tu-summary.json
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/wpa-result-bad-id.json b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/wpa-result-bad-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/wpa-result-bad-id.json
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/wpa-result-bad-id.json
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/wpa-result-empty.json b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/wpa-result-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/wpa-result-empty.json
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/wpa-result-empty.json
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/wpa-result-no-key.json b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/wpa-result-no-key.json
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/Inputs/wpa-result-no-key.json
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/Inputs/wpa-result-no-key.json
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/tu-summary-serialization.test b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/tu-summary-serialization.test
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/tu-summary-serialization.test
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/tu-summary-serialization.test
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/type-constrained-pointers.cpp b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/type-constrained-pointers.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/type-constrained-pointers.cpp
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/type-constrained-pointers.cpp
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-main.cpp b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-main.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-main.cpp
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-main.cpp
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-new-delete.cpp b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-new-delete.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-new-delete.cpp
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/unsafe-buffer-reachable-excludes-type-constrained-new-delete.cpp
diff --git a/clang/test/Analysis/Scalable/TypeConstrainedPointers/wpa-result-serialization.test b/clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/wpa-result-serialization.test
similarity index 100%
rename from clang/test/Analysis/Scalable/TypeConstrainedPointers/wpa-result-serialization.test
rename to clang/test/ScalableStaticAnalysis/TypeConstrainedPointers/wpa-result-serialization.test
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary-bad-element.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary-bad-element.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary-bad-element.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary-bad-element.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary-bad-ptr-level.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary-bad-ptr-level.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary-bad-ptr-level.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary-bad-ptr-level.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary-no-key.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary-no-key.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary-no-key.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary-no-key.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/tu-summary.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/tu-summary.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-bad-epls.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-bad-epls.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-bad-epls.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-bad-epls.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-bad-id.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-bad-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-bad-id.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-bad-id.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-empty.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-empty.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-empty.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-no-key.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-no-key.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-no-key.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-no-key.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-odd-count.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-odd-count.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result-odd-count.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result-odd-count.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result.json b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result.json
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/Inputs/wpa-result.json
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/Inputs/wpa-result.json
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/tu-summary-serialization.test b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/tu-summary-serialization.test
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/tu-summary-serialization.test
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/tu-summary-serialization.test
diff --git a/clang/test/Analysis/Scalable/UnsafeBufferUsage/wpa-result-serialization.test b/clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/wpa-result-serialization.test
similarity index 100%
rename from clang/test/Analysis/Scalable/UnsafeBufferUsage/wpa-result-serialization.test
rename to clang/test/ScalableStaticAnalysis/UnsafeBufferUsage/wpa-result-serialization.test
diff --git a/clang/test/Analysis/Scalable/call-graph-local-entities.cpp b/clang/test/ScalableStaticAnalysis/call-graph-local-entities.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/call-graph-local-entities.cpp
rename to clang/test/ScalableStaticAnalysis/call-graph-local-entities.cpp
diff --git a/clang/test/Analysis/Scalable/call-graph.cpp b/clang/test/ScalableStaticAnalysis/call-graph.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/call-graph.cpp
rename to clang/test/ScalableStaticAnalysis/call-graph.cpp
diff --git a/clang/test/Analysis/Scalable/cli-errors-compilation-unit-id.cpp b/clang/test/ScalableStaticAnalysis/cli-errors-compilation-unit-id.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/cli-errors-compilation-unit-id.cpp
rename to clang/test/ScalableStaticAnalysis/cli-errors-compilation-unit-id.cpp
diff --git a/clang/test/Analysis/Scalable/command-line-interface.cpp b/clang/test/ScalableStaticAnalysis/command-line-interface.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/command-line-interface.cpp
rename to clang/test/ScalableStaticAnalysis/command-line-interface.cpp
diff --git a/clang/test/Analysis/Scalable/downgradable-errors.cpp b/clang/test/ScalableStaticAnalysis/downgradable-errors.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/downgradable-errors.cpp
rename to clang/test/ScalableStaticAnalysis/downgradable-errors.cpp
diff --git a/clang/test/Analysis/Scalable/extraction-works-alongside-compilation.cpp b/clang/test/ScalableStaticAnalysis/extraction-works-alongside-compilation.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/extraction-works-alongside-compilation.cpp
rename to clang/test/ScalableStaticAnalysis/extraction-works-alongside-compilation.cpp
diff --git a/clang/test/Analysis/Scalable/help.cpp b/clang/test/ScalableStaticAnalysis/help.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/help.cpp
rename to clang/test/ScalableStaticAnalysis/help.cpp
diff --git a/clang/test/Analysis/Scalable/local-entities.cpp b/clang/test/ScalableStaticAnalysis/local-entities.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/local-entities.cpp
rename to clang/test/ScalableStaticAnalysis/local-entities.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/Inputs/empty-suite.json b/clang/test/ScalableStaticAnalysis/source-edit-generation/Inputs/empty-suite.json
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/Inputs/empty-suite.json
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/Inputs/empty-suite.json
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/Inputs/two-function-suite.json b/clang/test/ScalableStaticAnalysis/source-edit-generation/Inputs/two-function-suite.json
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/Inputs/two-function-suite.json
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/Inputs/two-function-suite.json
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/Plugins/CMakeLists.txt b/clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/CMakeLists.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/Plugins/CMakeLists.txt
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/CMakeLists.txt
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/Plugins/TestTransformationPlugin/CMakeLists.txt b/clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/TestTransformationPlugin/CMakeLists.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/Plugins/TestTransformationPlugin/CMakeLists.txt
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/TestTransformationPlugin/CMakeLists.txt
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/Plugins/TestTransformationPlugin/TestTransformation.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/TestTransformationPlugin/TestTransformation.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/Plugins/TestTransformationPlugin/TestTransformation.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/TestTransformationPlugin/TestTransformation.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/Plugins/lit.local.cfg b/clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/lit.local.cfg
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/Plugins/lit.local.cfg
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/Plugins/lit.local.cfg
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/cli-errors.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/cli-errors.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/cli-errors.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/cli-errors.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/coexistence.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/coexistence.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/coexistence.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/coexistence.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/cpp-bounded-buffers-replacements.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/cpp-bounded-buffers-replacements.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/cpp-bounded-buffers-replacements.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/cpp-bounded-buffers-replacements.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/downgradable-errors.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/downgradable-errors.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/downgradable-errors.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/downgradable-errors.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/happy-path.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/happy-path.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/happy-path.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/happy-path.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/lit.local.cfg b/clang/test/ScalableStaticAnalysis/source-edit-generation/lit.local.cfg
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/lit.local.cfg
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/lit.local.cfg
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/multi-tu-conflict-edits-merge.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/multi-tu-conflict-edits-merge.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/multi-tu-conflict-edits-merge.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/multi-tu-conflict-edits-merge.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/multi-tu-edits-merge.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/multi-tu-edits-merge.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/multi-tu-edits-merge.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/multi-tu-edits-merge.cpp
diff --git a/clang/test/Analysis/Scalable/source-edit-generation/write-failure.cpp b/clang/test/ScalableStaticAnalysis/source-edit-generation/write-failure.cpp
similarity index 100%
rename from clang/test/Analysis/Scalable/source-edit-generation/write-failure.cpp
rename to clang/test/ScalableStaticAnalysis/source-edit-generation/write-failure.cpp
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-badext.txt b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-badext.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-badext.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-badext.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-corrupt.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-corrupt.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-corrupt.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-corrupt.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-noext b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-noext
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-noext
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-noext
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-tags-only.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-tags-only.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-tags-only.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-tags-only.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-wrong-type.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-wrong-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu-wrong-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu-wrong-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Inputs/lu.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Inputs/lu.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/all.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/all.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/all.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/all.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/both.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/both.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/both.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/both.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/empty-pairs.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/empty-pairs.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/empty-pairs.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/empty-pairs.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/pairs.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/pairs.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/pairs.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/pairs.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/tags.json b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/tags.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/Outputs/tags.json
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/Outputs/tags.json
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/analyzer.test b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/analyzer.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/analyzer.test
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/analyzer.test
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/cli-errors.test b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/cli-errors.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/cli-errors.test
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/cli-errors.test
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/help.test b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/help.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/help.test
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/help.test
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/lit.local.cfg b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/lit.local.cfg
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/lit.local.cfg
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/lit.local.cfg
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/validation-errors-permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/validation-errors-permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/validation-errors-permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/validation-errors-permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/validation-errors.test b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/validation-errors.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/validation-errors.test
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/validation-errors.test
diff --git a/clang/test/Analysis/Scalable/ssaf-analyzer/version.test b/clang/test/ScalableStaticAnalysis/ssaf-analyzer/version.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-analyzer/version.test
rename to clang/test/ScalableStaticAnalysis/ssaf-analyzer/version.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-lu-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-lu-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-lu-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-lu-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-nonempty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-nonempty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-nonempty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-shared-library-nonempty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-nonempty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-nonempty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-nonempty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-multi-arch-static-library-nonempty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-static-library-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-static-library-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-static-library-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-static-library-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-tu-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-tu-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-tu-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-tu-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-wpa-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-wpa-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/rt-wpa-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/rt-wpa-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/unknown-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/unknown-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/Inputs/unknown-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/Inputs/unknown-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/round-trip.test b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/round-trip.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/round-trip.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/round-trip.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Artifact/top-level.test b/clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/top-level.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Artifact/top-level.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Artifact/top-level.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/invalid-direct-callee-element.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/invalid-direct-callee-element.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/invalid-direct-callee-element.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/invalid-direct-callee-element.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/invalid-direct-callee-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/invalid-direct-callee-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/invalid-direct-callee-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/invalid-direct-callee-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def-col.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def-col.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def-col.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def-col.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def-file.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def-file.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def-file.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def-file.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def-line.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def-line.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def-line.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def-line.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-def.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-def.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-direct-callees.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-direct-callees.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-direct-callees.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-direct-callees.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-pretty-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-pretty-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-pretty-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-pretty-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-virtual-callees.json b/clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-virtual-callees.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/Inputs/CallGraph/missing-virtual-callees.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/Inputs/CallGraph/missing-virtual-callees.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-entry-missing-data.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-entry-missing-data.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-entry-missing-data.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-entry-missing-data.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-entry-missing-summary-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-entry-missing-summary-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-entry-missing-summary-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-entry-missing-summary-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/data-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/data-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/duplicate-entity-id-in-data-map.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/duplicate-entity-id-in-data-map.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/duplicate-entity-id-in-data-map.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/duplicate-entity-id-in-data-map.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/duplicate-entity.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/duplicate-entity.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/duplicate-entity.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/duplicate-entity.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/duplicate-summary-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/duplicate-summary-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/duplicate-summary-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/duplicate-summary-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-data-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-data-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-data-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-data-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-summary.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-summary.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-summary.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-data-missing-entity-summary.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-id-not-uint64.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-id-not-uint64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-id-not-uint64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-id-not-uint64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-name-missing-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-name-missing-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-name-missing-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-name-missing-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-name-missing-suffix.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-name-missing-suffix.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-name-missing-suffix.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-name-missing-suffix.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-name-missing-usr.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-name-missing-usr.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/entity-name-missing-usr.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/entity-name-missing-usr.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-entry-id-not-uint64.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-entry-id-not-uint64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-entry-id-not-uint64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-entry-id-not-uint64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-entry-missing-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-entry-missing-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-entry-missing-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-entry-missing-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-entry-missing-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-entry-missing-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-entry-missing-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-entry-missing-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/id-table-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/id-table-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/invalid-syntax.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/invalid-syntax.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/invalid-syntax.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/invalid-syntax.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-duplicate-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-duplicate-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-duplicate-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-duplicate-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-id-not-uint64.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-id-not-uint64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-id-not-uint64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-id-not-uint64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-linkage-missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-linkage.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-linkage.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-linkage.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-entry-missing-linkage.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-extra-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-extra-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-extra-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-extra-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-missing-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-missing-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-missing-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-missing-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/linkage-table-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/linkage-table-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-invalid-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-invalid-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-invalid-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-invalid-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-missing-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/lu-namespace-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/lu-namespace-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-data.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-data.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-data.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-data.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-id-table.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-id-table.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-id-table.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-id-table.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-linkage-table.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-linkage-table.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-linkage-table.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-linkage-table.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-lu-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-lu-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-lu-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-lu-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-invalid-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-invalid-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-invalid-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-invalid-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-missing-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-missing-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-missing-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-missing-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-missing-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-missing-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-missing-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-missing-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/namespace-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/namespace-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/not-json-extension.txt b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/not-json-extension.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/not-json-extension.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/not-json-extension.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/not-normalized-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/not-normalized-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/not-normalized-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/not-normalized-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-invalid-first-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-invalid-first-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-invalid-first-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-invalid-first-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-invalid-pairs-field-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-invalid-pairs-field-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-invalid-pairs-field-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-invalid-pairs-field-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-invalid-second-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-invalid-second-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-invalid-second-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-invalid-second-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-missing-first-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-missing-first-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-missing-first-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-missing-first-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-missing-pairs-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-missing-pairs-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-missing-pairs-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-missing-pairs-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-missing-second-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-missing-second-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/pairs-missing-second-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/pairs-missing-second-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/read-entity-summary-no-format-info.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/read-entity-summary-no-format-info.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/read-entity-summary-no-format-info.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/read-entity-summary-no-format-info.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-empty-data-entry.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-empty-data-entry.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-empty-data-entry.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-empty-data-entry.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-empty-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-empty-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-empty-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-empty-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-external.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-external.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-external.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-external.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-internal.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-internal.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-internal.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-internal.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-multiple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-multiple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-multiple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-multiple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-none.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-none.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-linkage-none.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-linkage-none.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-multiple-namespace-elements.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-multiple-namespace-elements.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-multiple-namespace-elements.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-multiple-namespace-elements.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-single-namespace-element.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-single-namespace-element.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-single-namespace-element.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-single-namespace-element.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-two-summary-types.json b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-two-summary-types.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/Inputs/rt-two-summary-types.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/Inputs/rt-two-summary-types.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/id-table.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/id-table.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/id-table.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/id-table.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/io.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/io.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/io.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/io.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/linkage.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/linkage.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/linkage.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/linkage.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/round-trip.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/round-trip.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/round-trip.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/round-trip.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/summary-data.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/summary-data.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/summary-data.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/summary-data.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/LUSummary/top-level.test b/clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/top-level.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/LUSummary/top-level.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/LUSummary/top-level.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/duplicate-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/duplicate-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/duplicate-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/duplicate-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/invalid-syntax.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/invalid-syntax.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/invalid-syntax.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/invalid-syntax.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-empty-lu-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-empty-lu-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-empty-lu-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-empty-lu-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-inner-error.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-inner-error.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-inner-error.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-inner-error.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-namespace-mismatch.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-namespace-mismatch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-namespace-mismatch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-namespace-mismatch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/member-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/member-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/members-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/members-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/members-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/members-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/missing-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/missing-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/missing-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/missing-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/missing-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/missing-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/missing-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/missing-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/namespace-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/namespace-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/namespace-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/namespace-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/not-json-extension.txt b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/not-json-extension.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/not-json-extension.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/not-json-extension.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-empty-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-empty-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-empty-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-empty-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-multiple-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-multiple-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-multiple-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-multiple-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-nonempty-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-nonempty-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-nonempty-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-nonempty-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-single-member.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-single-member.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-single-member.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-single-member.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-three-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-three-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/rt-three-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/rt-three-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-members-input.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-members-input.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-members-input.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-members-input.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-three-members-input.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-three-members-input.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-three-members-input.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/Inputs/unsorted-three-members-input.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/io.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/io.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/io.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/io.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/round-trip.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/round-trip.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/round-trip.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/round-trip.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/top-level.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/top-level.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchSharedLibrary/top-level.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchSharedLibrary/top-level.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/duplicate-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/duplicate-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/duplicate-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/duplicate-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/invalid-syntax.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/invalid-syntax.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/invalid-syntax.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/invalid-syntax.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-inner-error.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-inner-error.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-inner-error.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-inner-error.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-namespace-mismatch.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-namespace-mismatch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-namespace-mismatch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-namespace-mismatch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/member-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/member-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/members-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/members-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/members-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/members-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/missing-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/missing-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/missing-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/missing-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/missing-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/missing-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/missing-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/missing-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-wrong-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-wrong-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-wrong-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/namespace-wrong-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/not-json-extension.txt b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/not-json-extension.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/not-json-extension.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/not-json-extension.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-empty-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-empty-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-empty-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-empty-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-multiple-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-multiple-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-multiple-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-multiple-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-nonempty-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-nonempty-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-nonempty-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-nonempty-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-single-member.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-single-member.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-single-member.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-single-member.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-three-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-three-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/rt-three-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/rt-three-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-members-input.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-members-input.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-members-input.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-members-input.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-three-members-input.json b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-three-members-input.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-three-members-input.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/Inputs/unsorted-three-members-input.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/io.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/io.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/io.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/io.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/round-trip.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/round-trip.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/round-trip.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/round-trip.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/top-level.test b/clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/top-level.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/MultiArchStaticLibrary/top-level.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/MultiArchStaticLibrary/top-level.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/duplicate-member.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/duplicate-member.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/duplicate-member.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/duplicate-member.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/invalid-syntax.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/invalid-syntax.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/invalid-syntax.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/invalid-syntax.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-inner-error.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-inner-error.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-inner-error.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-inner-error.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-triple-mismatch.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-triple-mismatch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/member-triple-mismatch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/member-triple-mismatch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/members-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/members-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/members-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/members-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-invalid-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-invalid-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-invalid-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-invalid-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-missing-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-missing-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-missing-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-missing-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-missing-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-missing-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-missing-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-missing-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-wrong-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-wrong-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/namespace-wrong-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/namespace-wrong-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/not-json-extension.txt b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/not-json-extension.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/not-json-extension.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/not-json-extension.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/not-normalized-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/not-normalized-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/not-normalized-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/not-normalized-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/rt-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/rt-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/rt-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/rt-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/rt-multiple-members.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/rt-multiple-members.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/rt-multiple-members.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/rt-multiple-members.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/rt-single-member.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/rt-single-member.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/rt-single-member.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/rt-single-member.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/unsorted-members-input.json b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/unsorted-members-input.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/Inputs/unsorted-members-input.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/Inputs/unsorted-members-input.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/io.test b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/io.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/io.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/io.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/round-trip.test b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/round-trip.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/round-trip.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/round-trip.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/top-level.test b/clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/top-level.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/StaticLibrary/top-level.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/StaticLibrary/top-level.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-entry-missing-data.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-entry-missing-data.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-entry-missing-data.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-entry-missing-data.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-entry-missing-summary-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-entry-missing-summary-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-entry-missing-summary-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-entry-missing-summary-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/data-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/data-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/duplicate-entity-id-in-data-map.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/duplicate-entity-id-in-data-map.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/duplicate-entity-id-in-data-map.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/duplicate-entity-id-in-data-map.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/duplicate-entity.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/duplicate-entity.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/duplicate-entity.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/duplicate-entity.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/duplicate-summary-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/duplicate-summary-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/duplicate-summary-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/duplicate-summary-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-data-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-data-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-data-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-data-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-summary.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-summary.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-summary.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-data-missing-entity-summary.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-id-not-uint64.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-id-not-uint64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-id-not-uint64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-id-not-uint64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-name-missing-suffix.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-name-missing-suffix.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-name-missing-suffix.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-name-missing-suffix.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-name-missing-usr.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-name-missing-usr.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/entity-name-missing-usr.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/entity-name-missing-usr.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-entry-id-not-uint64.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-entry-id-not-uint64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-entry-id-not-uint64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-entry-id-not-uint64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-entry-missing-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-entry-missing-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-entry-missing-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-entry-missing-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-entry-missing-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-entry-missing-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-entry-missing-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-entry-missing-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/id-table-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/id-table-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/invalid-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/invalid-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/invalid-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/invalid-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/invalid-syntax.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/invalid-syntax.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/invalid-syntax.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/invalid-syntax.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-duplicate-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-duplicate-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-duplicate-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-duplicate-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-id-not-uint64.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-id-not-uint64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-id-not-uint64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-id-not-uint64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-invalid-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-linkage-missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-linkage.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-linkage.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-linkage.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-entry-missing-linkage.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-extra-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-extra-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-extra-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-extra-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-missing-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-missing-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-missing-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-missing-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/linkage-table-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/linkage-table-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-data.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-data.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-data.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-data.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-id-table.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-id-table.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-id-table.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-id-table.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-kind.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-kind.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-kind.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-kind.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-linkage-table.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-linkage-table.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-linkage-table.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-linkage-table.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-tu-namespace.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-tu-namespace.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-tu-namespace.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-tu-namespace.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/not-json-extension.txt b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/not-json-extension.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/not-json-extension.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/not-json-extension.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/not-normalized-target-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/not-normalized-target-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/not-normalized-target-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/not-normalized-target-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-element-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-element-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-element-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-element-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-invalid-first-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-invalid-first-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-invalid-first-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-invalid-first-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-invalid-pairs-field-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-invalid-pairs-field-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-invalid-pairs-field-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-invalid-pairs-field-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-invalid-second-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-invalid-second-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-invalid-second-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-invalid-second-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-missing-first-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-missing-first-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-missing-first-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-missing-first-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-missing-pairs-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-missing-pairs-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-missing-pairs-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-missing-pairs-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-missing-second-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-missing-second-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/pairs-missing-second-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/pairs-missing-second-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/read-entity-summary-no-format-info.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/read-entity-summary-no-format-info.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/read-entity-summary-no-format-info.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/read-entity-summary-no-format-info.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-empty-data-entry.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-empty-data-entry.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-empty-data-entry.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-empty-data-entry.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-link-unit.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-link-unit.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-link-unit.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-link-unit.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-external.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-external.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-external.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-external.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-internal.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-internal.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-internal.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-internal.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-multiple.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-multiple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-multiple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-multiple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-none.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-none.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-linkage-none.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-linkage-none.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-two-summary-types.json b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-two-summary-types.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/Inputs/rt-two-summary-types.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/Inputs/rt-two-summary-types.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/id-table.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/id-table.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/id-table.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/id-table.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/io.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/io.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/io.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/io.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/linkage.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/linkage.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/linkage.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/linkage.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/round-trip.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/round-trip.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/round-trip.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/round-trip.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/summary-data.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/summary-data.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/summary-data.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/summary-data.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/TUSummary/top-level.test b/clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/top-level.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/TUSummary/top-level.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/TUSummary/top-level.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/duplicate-analysis-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/duplicate-analysis-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/duplicate-analysis-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/duplicate-analysis-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/mismatched-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/mismatched-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/mismatched-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/mismatched-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/missing-results.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/missing-results.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/missing-results.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/missing-results.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/missing-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/missing-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/missing-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/missing-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-bad-count.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-bad-count.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-bad-count.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-bad-count.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-bad-pair-counts.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-bad-pair-counts.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-bad-pair-counts.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-bad-pair-counts.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-entry-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-entry-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-entry-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-entry-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-missing-entity-id.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-missing-entity-id.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/pairs-result-missing-entity-id.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/pairs-result-missing-entity-id.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-analysis-name-not-string.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-analysis-name-not-string.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-analysis-name-not-string.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-analysis-name-not-string.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-deserializer-error.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-deserializer-error.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-deserializer-error.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-deserializer-error.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-missing-analysis-name.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-missing-analysis-name.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-missing-analysis-name.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-missing-analysis-name.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-missing-result-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-missing-result-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-missing-result-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-missing-result-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-no-format-info.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-no-format-info.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-no-format-info.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-no-format-info.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-result-not-object.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-result-not-object.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/result-entry-result-not-object.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/result-entry-result-not-object.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/results-not-array.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/results-not-array.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/results-not-array.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/results-not-array.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/rt-all-results.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/rt-all-results.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/rt-all-results.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/rt-all-results.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/rt-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/rt-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/rt-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/rt-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/rt-multiple-results.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/rt-multiple-results.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/rt-multiple-results.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/rt-multiple-results.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tags-result-bad-element.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tags-result-bad-element.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tags-result-bad-element.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tags-result-bad-element.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tags-result-missing-tags.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tags-result-missing-tags.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tags-result-missing-tags.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tags-result-missing-tags.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tagspairs-result-bad-field-type.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tagspairs-result-bad-field-type.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tagspairs-result-bad-field-type.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tagspairs-result-bad-field-type.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tagspairs-result-missing-field.json b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tagspairs-result-missing-field.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/Inputs/tagspairs-result-missing-field.json
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/Inputs/tagspairs-result-missing-field.json
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/with-plugin.test b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/with-plugin.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/with-plugin.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/with-plugin.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/WPASuite/without-plugin.test b/clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/without-plugin.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/WPASuite/without-plugin.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/WPASuite/without-plugin.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/call-graph-invalid-json-format.test b/clang/test/ScalableStaticAnalysis/ssaf-format/call-graph-invalid-json-format.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/call-graph-invalid-json-format.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/call-graph-invalid-json-format.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/list.test b/clang/test/ScalableStaticAnalysis/ssaf-format/list.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/list.test
rename to clang/test/ScalableStaticAnalysis/ssaf-format/list.test
diff --git a/clang/test/Analysis/Scalable/ssaf-format/lit.local.cfg b/clang/test/ScalableStaticAnalysis/ssaf-format/lit.local.cfg
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-format/lit.local.cfg
rename to clang/test/ScalableStaticAnalysis/ssaf-format/lit.local.cfg
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/bad-artifact.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/bad-artifact.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/bad-artifact.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/bad-artifact.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-aarch64.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-aarch64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-aarch64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-aarch64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-arm64.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-arm64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-arm64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-arm64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-ios-versioned.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-ios-versioned.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-ios-versioned.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-ios-versioned.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-ios.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-ios.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-ios.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-ios.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-otherns.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-otherns.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-otherns.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-otherns.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-x86_64.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-x86_64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lib-x86_64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lib-x86_64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libfoo-2arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libfoo-2arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libfoo-2arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libfoo-2arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libmulti-empty-wrapper.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libmulti-empty-wrapper.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libmulti-empty-wrapper.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libmulti-empty-wrapper.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libmulti-multiarch-wrongns.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libmulti-multiarch-wrongns.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libmulti-multiarch-wrongns.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libmulti-multiarch-wrongns.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libord-reversed.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libord-reversed.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libord-reversed.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libord-reversed.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libshared-empty-wrapper.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libshared-empty-wrapper.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libshared-empty-wrapper.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libshared-empty-wrapper.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libshared-multiarch-wrongns.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libshared-multiarch-wrongns.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libshared-multiarch-wrongns.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libshared-multiarch-wrongns.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/libtwo-2arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libtwo-2arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/libtwo-2arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/libtwo-2arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-arm64.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-arm64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-arm64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-arm64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-ios-libfoo.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-ios-libfoo.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-ios-libfoo.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-ios-libfoo.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-macosx-libfoo.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-macosx-libfoo.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-macosx-libfoo.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-macosx-libfoo.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-otherns.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-otherns.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-otherns.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-otherns.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-x86_64.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-x86_64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/lu-x86_64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/lu-x86_64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-1.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-1.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-1.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-1.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-2.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-2.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-2.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-2.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-badext.txt b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-badext.txt
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-badext.txt
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-badext.txt
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-invalid-entity-id-multikey.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-invalid-entity-id-multikey.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-invalid-entity-id-multikey.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-invalid-entity-id-multikey.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-invalid-entity-id-ref.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-invalid-entity-id-ref.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-invalid-entity-id-ref.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-invalid-entity-id-ref.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-invalid-entity-id-value.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-invalid-entity-id-value.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-invalid-entity-id-value.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-invalid-entity-id-value.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-linux.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-linux.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-linux.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-linux.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-malformed.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-malformed.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-malformed.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-malformed.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-missing-fields.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-missing-fields.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-missing-fields.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-missing-fields.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-noext b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-noext
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-noext
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-noext
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-x86_64.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-x86_64.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Inputs/tu-x86_64.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Inputs/tu-x86_64.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/lib-explicit-triple.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lib-explicit-triple.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/lib-explicit-triple.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lib-explicit-triple.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/lib-namespace-override.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lib-namespace-override.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/lib-namespace-override.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lib-namespace-override.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libempty.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libempty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libempty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libempty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libfoo-3arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libfoo-3arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libfoo-3arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libfoo-3arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libmulti-1arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libmulti-1arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libmulti-1arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libmulti-1arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libmulti-2arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libmulti-2arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libmulti-2arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libmulti-2arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libmulti-3arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libmulti-3arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libmulti-3arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libmulti-3arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libshared-1arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libshared-1arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libshared-1arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libshared-1arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libshared-2arch.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libshared-2arch.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libshared-2arch.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libshared-2arch.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/libtwo.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libtwo.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/libtwo.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/libtwo.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-1+2.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-1+2.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-1+2.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-1+2.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-1.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-1.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-1.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-1.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-2.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-2.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-2.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-2.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-empty.json b/clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-empty.json
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/Outputs/lu-empty.json
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/Outputs/lu-empty.json
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/cli.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/cli.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/cli.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/cli.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/help.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/help.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/help.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/help.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/io.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/io.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/io.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/io.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/linking-errors.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/linking-errors.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/linking-errors.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/linking-errors.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/linking.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/linking.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/linking.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/linking.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/multi-arch-create.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-create.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/multi-arch-create.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-create.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/multi-arch-help.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-help.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/multi-arch-help.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-help.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/multi-arch-time.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-time.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/multi-arch-time.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-time.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/multi-arch-verbose.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-verbose.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/multi-arch-verbose.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/multi-arch-verbose.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/static-library-create.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-create.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/static-library-create.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-create.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/static-library-help.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-help.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/static-library-help.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-help.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/static-library-time.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-time.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/static-library-time.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-time.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/static-library-verbose.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-verbose.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/static-library-verbose.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/static-library-verbose.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/time.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/time.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/time.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/time.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/validation-errors-permissions.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/validation-errors-permissions.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/validation-errors-permissions.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/validation-errors-permissions.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/validation-errors.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/validation-errors.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/validation-errors.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/validation-errors.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/verbose.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/verbose.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/verbose.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/verbose.test
diff --git a/clang/test/Analysis/Scalable/ssaf-linker/version.test b/clang/test/ScalableStaticAnalysis/ssaf-linker/version.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-linker/version.test
rename to clang/test/ScalableStaticAnalysis/ssaf-linker/version.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/conflict-text-q-offset-30.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/conflict-text-q-offset-30.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/conflict-text-q-offset-30.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/conflict-text-q-offset-30.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/conflict-text-y.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/conflict-text-y.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/conflict-text-y.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/conflict-text-y.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/conflict-text-z.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/conflict-text-z.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/conflict-text-z.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/conflict-text-z.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/empty.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/empty.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/empty.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/empty.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/internal-conflict.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/internal-conflict.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/internal-conflict.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/internal-conflict.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/non-overlapping-a.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/non-overlapping-a.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/non-overlapping-a.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/non-overlapping-a.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/non-overlapping-b.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/non-overlapping-b.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/non-overlapping-b.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/non-overlapping-b.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/range-overlap-cluster-of-three.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/range-overlap-cluster-of-three.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/range-overlap-cluster-of-three.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/range-overlap-cluster-of-three.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/range-overlap-distinct-offsets.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/range-overlap-distinct-offsets.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/range-overlap-distinct-offsets.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/range-overlap-distinct-offsets.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-non-zero-length-replacement.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-non-zero-length-replacement.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-non-zero-length-replacement.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-non-zero-length-replacement.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-replacement-offset-30.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-replacement-offset-30.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-replacement-offset-30.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-replacement-offset-30.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-replacement.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-replacement.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-replacement.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-replacement.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-zero-length-insert.tpl.yaml b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-zero-length-insert.tpl.yaml
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/Inputs/single-zero-length-insert.tpl.yaml
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/Inputs/single-zero-length-insert.tpl.yaml
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/bad-args.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/bad-args.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/bad-args.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/bad-args.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/conflict-preserves-unrelated-zero-length-insert.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/conflict-preserves-unrelated-zero-length-insert.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/conflict-preserves-unrelated-zero-length-insert.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/conflict-preserves-unrelated-zero-length-insert.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/conflict-stderr-without-sarif.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/conflict-stderr-without-sarif.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/conflict-stderr-without-sarif.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/conflict-stderr-without-sarif.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/empty-input.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/empty-input.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/empty-input.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/empty-input.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/help.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/help.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/help.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/help.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/identical-overlap.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/identical-overlap.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/identical-overlap.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/identical-overlap.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/malformed-yaml.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/malformed-yaml.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/malformed-yaml.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/malformed-yaml.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/no-conflict-sarif-empty.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/no-conflict-sarif-empty.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/no-conflict-sarif-empty.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/no-conflict-sarif-empty.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/non-overlapping.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/non-overlapping.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/non-overlapping.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/non-overlapping.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/order-invariance.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/order-invariance.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/order-invariance.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/order-invariance.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/parent-dir-missing.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/parent-dir-missing.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/parent-dir-missing.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/parent-dir-missing.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/pre-dedup-identical-non-zero-length.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/pre-dedup-identical-non-zero-length.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/pre-dedup-identical-non-zero-length.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/pre-dedup-identical-non-zero-length.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/pre-dedup-zero-length-insert.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/pre-dedup-zero-length-insert.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/pre-dedup-zero-length-insert.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/pre-dedup-zero-length-insert.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/range-overlap-cluster-of-three-sarif.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/range-overlap-cluster-of-three-sarif.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/range-overlap-cluster-of-three-sarif.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/range-overlap-cluster-of-three-sarif.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/range-overlap-conflict-sarif.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/range-overlap-conflict-sarif.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/range-overlap-conflict-sarif.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/range-overlap-conflict-sarif.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/sarif-out-parent-dir-missing.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/sarif-out-parent-dir-missing.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/sarif-out-parent-dir-missing.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/sarif-out-parent-dir-missing.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/single-input-internal-conflict.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/single-input-internal-conflict.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/single-input-internal-conflict.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/single-input-internal-conflict.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/single-input.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/single-input.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/single-input.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/single-input.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/three-input-all-conflict.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/three-input-all-conflict.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/three-input-all-conflict.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/three-input-all-conflict.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/three-input-conflict-sarif.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/three-input-conflict-sarif.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/three-input-conflict-sarif.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/three-input-conflict-sarif.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/two-conflicts-different-offsets-sarif.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/two-conflicts-different-offsets-sarif.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/two-conflicts-different-offsets-sarif.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/two-conflicts-different-offsets-sarif.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/two-input-conflict-sarif.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/two-input-conflict-sarif.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/two-input-conflict-sarif.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/two-input-conflict-sarif.test
diff --git a/clang/test/Analysis/Scalable/ssaf-src-edit-merge/two-input-conflict.test b/clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/two-input-conflict.test
similarity index 100%
rename from clang/test/Analysis/Scalable/ssaf-src-edit-merge/two-input-conflict.test
rename to clang/test/ScalableStaticAnalysis/ssaf-src-edit-merge/two-input-conflict.test
diff --git a/clang/unittests/ScalableStaticAnalysis/CMakeLists.txt b/clang/unittests/ScalableStaticAnalysis/CMakeLists.txt
index a8195cedefd141..f01f62d5caf558 100644
--- a/clang/unittests/ScalableStaticAnalysis/CMakeLists.txt
+++ b/clang/unittests/ScalableStaticAnalysis/CMakeLists.txt
@@ -64,3 +64,12 @@ add_distinct_clang_unittest(ClangScalableAnalysisTests
Support
TargetParser
)
+
+add_custom_target(check-clang-scalablestaticanalysis-unit
+ COMMAND ClangScalableAnalysisTests
+ DEPENDS ClangScalableAnalysisTests
+ COMMENT "Running the Scalable Static Analysis Framework unit tests"
+ USES_TERMINAL
+ )
+set_target_properties(check-clang-scalablestaticanalysis-unit
+ PROPERTIES FOLDER "Clang tests/Suites")
>From 3f3bd3f9e0bc3fdb24219fbfebdc2ac4edfc3e5c Mon Sep 17 00:00:00 2001
From: actink <actink at 163.com>
Date: Sat, 21 Nov 2026 18:33:59 +0800
Subject: [PATCH 06/35] [PowerPC] Add a precommit test for MTCRF CR restore
scheduling (#242760)
This adds a regression test that pins the current (incorrect) order of
the mtcrf that restores the non-volatile CR fields relative to reads of
the CR bits it writes. A follow-up patch will model the CR fields as
defs and update the CHECK lines; the FIXME marks the expected change.
See #242250
---
llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll | 77 +++++++++++++++++++
1 file changed, 77 insertions(+)
create mode 100644 llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll
diff --git a/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll b/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll
new file mode 100644
index 00000000000000..8a0dd5d8e4e1da
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll
@@ -0,0 +1,77 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr8 -O2 -verify-machineinstrs < %s | FileCheck %s
+;
+; The epilogue restores the non-volatile CR fields with mtcrf. The post-RA
+; scheduler must not move that restore above reads of the CR bits it writes.
+;
+; FIXME: MTCRF does not model the CR fields it writes, so the restore is
+; currently hoisted above some CR2/CR3 reads. Update this test once MTCRF
+; models its defs. See https://github.com/llvm/llvm-project/issues/242250.
+
+define signext i32 @restore_cr_after_uses(i128 %value, i1 %cond) {
+; CHECK-LABEL: restore_cr_after_uses:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: addic 6, 3, 1
+; CHECK-NEXT: andi. 5, 5, 1
+; CHECK-NEXT: li 5, 0
+; CHECK-NEXT: addze 7, 4
+; CHECK-NEXT: cmpldi 5, 6, 1
+; CHECK-NEXT: li 6, 0
+; CHECK-NEXT: cmpldi 1, 7, 0
+; CHECK-NEXT: cmpdi 6, 7, 1
+; CHECK-NEXT: bc 12, 1, .LBB0_2
+; CHECK-NEXT: # %bb.1: # %compare
+; CHECK-NEXT: crandc 20, 24, 6
+; CHECK-NEXT: crandc 22, 6, 21
+; CHECK-NEXT: li 6, 1
+; CHECK-NEXT: crnor 20, 22, 20
+; CHECK-NEXT: isel 6, 0, 6, 20
+; CHECK-NEXT: .LBB0_2: # %merge
+; CHECK-NEXT: clrldi 6, 6, 32
+; CHECK-NEXT: mfcr 12
+; CHECK-NEXT: cmpldi 7, 4, 0
+; CHECK-NEXT: cmpdi 2, 4, 0
+; CHECK-NEXT: crandc 21, 6, 21
+; CHECK-NEXT: stw 12, 8(1)
+; CHECK-NEXT: cmpld 3, 3, 6
+; CHECK-NEXT: li 3, 1
+; CHECK-NEXT: crandc 20, 8, 30
+; CHECK-NEXT: lwz 12, 8(1)
+; CHECK-NEXT: mtcrf 48, 12
+; CHECK-NEXT: crand 22, 30, 12
+; CHECK-NEXT: crnor 20, 22, 20
+; CHECK-NEXT: isel 4, 0, 3, 20
+; CHECK-NEXT: crandc 20, 24, 6
+; CHECK-NEXT: crnor 20, 21, 20
+; CHECK-NEXT: isel 6, 5, 3, 20
+; CHECK-NEXT: iselgt 3, 3, 5
+; CHECK-NEXT: add 4, 4, 6
+; CHECK-NEXT: addi 4, 4, 1
+; CHECK-NEXT: or 3, 4, 3
+; CHECK-NEXT: clrldi 3, 3, 32
+; CHECK-NEXT: blr
+entry:
+ %value.plus.one = add i128 %value, 1
+ br i1 %cond, label %early.exit, label %compare
+
+early.exit: ; preds = %entry
+ br label %merge
+
+compare: ; preds = %entry
+ %cmp.le = icmp sle i128 %value.plus.one, 1
+ %cmp.le.ext = zext i1 %cmp.le to i32
+ br label %merge
+
+merge: ; preds = %compare, %early.exit
+ %partial = phi i32 [ %cmp.le.ext, %compare ], [ 0, %early.exit ]
+ %partial.ext = zext i32 %partial to i128
+ %cmp.value = icmp slt i128 %value, %partial.ext
+ %cmp.value.ext = zext i1 %cmp.value to i32
+ %sum = add i32 1, %cmp.value.ext
+ %cmp.le.again = icmp sle i128 %value.plus.one, 1
+ %cmp.le.again.ext = zext i1 %cmp.le.again to i32
+ %sum.plus = add i32 %sum, %cmp.le.again.ext
+ %cond.ext = zext i1 %cond to i32
+ %result = or i32 %sum.plus, %cond.ext
+ ret i32 %result
+}
>From df0be3ef56b8161e18b222201833c9b668f27707 Mon Sep 17 00:00:00 2001
From: Matthew <matthew.arda at gmail.com>
Date: Sat, 21 Nov 2026 10:48:38 +0000
Subject: [PATCH 07/35] [clang-tidy] Add SARIF output to clang-tidy (#219182)
[clang-tidy] Add SARIF output for clang-tidy diagnostics.
Clang-tidy is a static analysis framework, and should therefore be able
to output results in the Static Analysis Results Interchange Format
(SARIF). This brings clang-tidy in line with clang and
clang-static-analyzer.
Adds an experimental '--experimental-sarif-export=<file>' option to
clang-tidy that writes diagnostics in SARIF format.
Each diagnostic is reported as a SARIF result with its check name,
severity, and location. Notes are emitted as related locations with
their messages, while notes from clang-analyzer checks are emitted as
thread flows so the analyzer path can be followed in SARIF viewers.
To support this, the shared SARIF writer in clang now skips locations
that are invalid or not in a file instead of asserting, allows a message
on related locations, and fixes artifact indices not matching their
position in the artifacts array. The diagnostic level helpers in
SARIFDiagnostic are made public so clang-tidy reports severities
consistently with clang.
Closes #55647
---------
Co-authored-by: Jones, Matthew (DI SW SIM FRWK PLAT SIMFW HPC) <jones.matthew at siemens.com>
---
clang-tools-extra/clang-tidy/ClangTidy.cpp | 140 +++++++-
clang-tools-extra/clang-tidy/ClangTidy.h | 3 +-
.../ClangTidyDiagnosticConsumer.cpp | 38 +--
.../clang-tidy/tool/ClangTidyMain.cpp | 33 +-
clang-tools-extra/docs/ReleaseNotes.md | 4 +
clang-tools-extra/docs/clang-tidy/index.rst | 3 +
.../export-SARIF-conflicting-output.cpp | 8 +
.../export-SARIF-diagnostics-EOF.cpp | 76 +++++
...t-SARIF-diagnostics-macro-thread-flows.cpp | 175 ++++++++++
.../export-SARIF-diagnostics-stdout.cpp | 302 +++++++++++++++++
.../export-SARIF-diagnostics.cpp | 318 ++++++++++++++++++
.../export-SARIF-no-location.cpp | 104 ++++++
clang/include/clang/Basic/Sarif.h | 24 +-
.../include/clang/Frontend/SARIFDiagnostic.h | 16 +-
clang/lib/Basic/Sarif.cpp | 84 +++--
clang/lib/Frontend/SARIFDiagnostic.cpp | 66 ++--
clang/unittests/Basic/SarifTest.cpp | 19 +-
17 files changed, 1311 insertions(+), 102 deletions(-)
create mode 100644 clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-conflicting-output.cpp
create mode 100644 clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-EOF.cpp
create mode 100644 clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-macro-thread-flows.cpp
create mode 100644 clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-stdout.cpp
create mode 100644 clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics.cpp
create mode 100644 clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-no-location.cpp
diff --git a/clang-tools-extra/clang-tidy/ClangTidy.cpp b/clang-tools-extra/clang-tidy/ClangTidy.cpp
index d065bb2c4bcfcd..d1ef8de8b4a9cc 100644
--- a/clang-tools-extra/clang-tidy/ClangTidy.cpp
+++ b/clang-tools-extra/clang-tidy/ClangTidy.cpp
@@ -24,10 +24,12 @@
#include "clang/AST/ASTConsumer.h"
#include "clang/ASTMatchers/ASTMatchFinder.h"
#include "clang/Basic/DiagnosticFrontend.h"
+#include "clang/Basic/Sarif.h"
#include "clang/Format/Format.h"
#include "clang/Frontend/ASTConsumers.h"
#include "clang/Frontend/CompilerInstance.h"
#include "clang/Frontend/MultiplexConsumer.h"
+#include "clang/Frontend/SARIFDiagnostic.h"
#include "clang/Frontend/TextDiagnosticPrinter.h"
#include "clang/Lex/Preprocessor.h"
#include "clang/Lex/PreprocessorOptions.h"
@@ -36,7 +38,11 @@
#include "clang/Tooling/DiagnosticsYaml.h" // IWYU pragma: keep
#include "clang/Tooling/Refactoring.h"
#include "clang/Tooling/Tooling.h"
+#include "llvm/Support/Error.h"
+#include "llvm/Support/FileSystem.h"
+#include "llvm/Support/FormatVariadic.h"
#include "llvm/Support/Process.h"
+#include "llvm/Support/raw_ostream.h"
#include <memory>
#include <utility>
@@ -101,11 +107,22 @@ class AnalyzerDiagnosticConsumer : public ento::PathDiagnosticConsumer {
class ErrorReporter {
public:
ErrorReporter(ClangTidyContext &Context, FixBehaviour ApplyFixes,
- llvm::IntrusiveRefCntPtr<llvm::vfs::FileSystem> BaseFS)
+ llvm::IntrusiveRefCntPtr<llvm::vfs::FileSystem> BaseFS,
+ StringRef SarifExportPath)
: Files(FileSystemOptions(), std::move(BaseFS)),
- DiagPrinter(new TextDiagnosticPrinter(llvm::outs(), DiagOpts)),
+ DiagPrinter(
+ SarifExportPath == "-"
+ ? static_cast<DiagnosticConsumer *>(new IgnoringDiagConsumer())
+ : new TextDiagnosticPrinter(llvm::outs(), DiagOpts)),
Diags(DiagnosticIDs::create(), DiagOpts, DiagPrinter),
- SourceMgr(Diags, Files), Context(Context), ApplyFixes(ApplyFixes) {
+ SourceMgr(Diags, Files), Context(Context), ApplyFixes(ApplyFixes),
+ SarifExportPath(SarifExportPath) {
+ if (!SarifExportPath.empty()) {
+ SarifWriter.emplace(SourceMgr);
+ SarifWriter->createRun("clang-tidy", "clang-tidy", CLANG_VERSION_STRING,
+ "https://clang.llvm.org/extra/clang-tidy/");
+ }
+
DiagOpts.setShowColors(Context.getOptions().UseColor.value_or(
llvm::sys::Process::StandardOutHasColors())
? ShowColorsKind::On
@@ -195,6 +212,90 @@ class ErrorReporter {
}
for (const auto &Note : Error.Notes)
reportNote(Note);
+
+ if (SarifWriter)
+ exportSarifResult(Error, Loc);
+ }
+
+ CharSourceRange getNoteRange(const tooling::DiagnosticMessage &Note) {
+ const SourceLocation Loc = getLocation(Note.FilePath, Note.FileOffset);
+ return Loc.isValid() ? getOffsetRange(SourceMgr, Loc) : CharSourceRange();
+ }
+
+ SmallVector<ThreadFlow> createNotesThreadFlows(const ClangTidyError &Error) {
+ SmallVector<ThreadFlow> Flows;
+ Flows.reserve(Error.Notes.size());
+ for (const tooling::DiagnosticMessage &Note : Error.Notes) {
+ Flows.push_back(ThreadFlow::create()
+ .setRange(getNoteRange(Note))
+ .setMessage(Note.Message));
+ }
+ return Flows;
+ }
+
+ void exportSarifResult(const ClangTidyError &Error,
+ const SourceLocation Loc) {
+ assert(SarifWriter && "SARIF writer must be initialized");
+
+ const auto [RuleIndexEntryIt, Inserted] =
+ SarifRuleIdx.try_emplace(Error.DiagnosticName, 0);
+ size_t &RuleIndex = RuleIndexEntryIt->second;
+
+ const DiagnosticsEngine::Level EffectiveLevel =
+ Error.IsWarningAsError
+ ? DiagnosticsEngine::Error
+ : static_cast<DiagnosticsEngine::Level>(Error.DiagLevel);
+
+ if (Inserted) {
+ const StringRef Name = Error.DiagnosticName;
+ SarifRule Rule = SarifRule::create().setRuleId(Name).setName(Name);
+ Rule = addDiagnosticLevelToRule(Rule, EffectiveLevel);
+ RuleIndex = SarifWriter->createRule(Rule);
+ }
+
+ SarifResult Result =
+ SarifResult::create(RuleIndex)
+ .setDiagnosticMessage(Error.Message.Message)
+ .setDiagnosticLevel(getSarifResultLevel(EffectiveLevel));
+ // TODO: Extend SarifDocumentWriter to add an annotations property that can
+ // represent highlight ranges, when both a valid location and highlight
+ // ranges are present.
+ Result =
+ (Loc.isValid() ? Result.addLocations(getOffsetRange(SourceMgr, Loc))
+ : Result.addLocations(getResultRanges(Error)));
+
+ // Same-location, same-message diagnostics from other checks are merged,
+ // with the first alphabetically as the check name, not necessarily the
+ // canon name.
+ if (!Error.EnabledDiagnosticAliases.empty())
+ Result = Result.addProperty(
+ "enabledAliases", llvm::join(Error.EnabledDiagnosticAliases, ","));
+
+ // TODO: CSA stores fullDescription from the HELPTEXT field in Checkers.inc,
+ // strip clang-analyzer- prefix then populate fullDescription with this.
+ // Also, add helpUri with this for in-tree checks, pointing to their
+ // documentation pages.
+ if (StringRef(Error.DiagnosticName).starts_with("clang-analyzer-"))
+ Result = Result.setThreadFlows(createNotesThreadFlows(Error));
+ else
+ for (const tooling::DiagnosticMessage &Note : Error.Notes)
+ Result = Result.addRelatedLocations(getNoteRange(Note), Note.Message);
+ SarifWriter->appendResult(Result);
+ }
+
+ // SARIF regions need a character range: use 1 char at Loc, or 0 at EOF.
+ CharSourceRange getOffsetRange(SourceManager &SM, SourceLocation Loc) {
+ const auto [FileId, Offset] = SM.getDecomposedLoc(Loc);
+ return CharSourceRange::getCharRange(
+ Loc, Loc.getLocWithOffset(Offset < SM.getFileIDSize(FileId) ? 1 : 0));
+ }
+
+ SmallVector<CharSourceRange> getResultRanges(const ClangTidyError &Error) {
+ SmallVector<CharSourceRange> Ranges;
+ Ranges.reserve(Error.Message.Ranges.size());
+ for (const FileByteRange &FBR : Error.Message.Ranges)
+ Ranges.push_back(getRange(FBR));
+ return Ranges;
}
void finish() {
@@ -254,6 +355,8 @@ class ErrorReporter {
if (OriginalCWD)
VFS.setCurrentWorkingDirectory(*OriginalCWD);
}
+ if (SarifWriter)
+ finalizeSarif();
}
unsigned getWarningsAsErrorsCount() const { return WarningsAsErrors; }
@@ -271,6 +374,24 @@ class ErrorReporter {
return SourceMgr.getLocForStartOfFile(ID).getLocWithOffset(Offset);
}
+ void finalizeSarif() {
+ assert(SarifWriter && "SARIF writer must be initialized");
+ const llvm::json::Value Document = SarifWriter->createDocument();
+ if (SarifExportPath == "-") {
+ llvm::outs() << llvm::formatv("{0:2}", Document);
+ return;
+ }
+ // Reopened here as the file may have changed during the analysis.
+ std::error_code EC;
+ llvm::raw_fd_ostream SarifOS(SarifExportPath, EC, llvm::sys::fs::OF_Text);
+ if (EC) {
+ llvm::errs() << "Error opening SARIF export file '" << SarifExportPath
+ << "': " << EC.message() << "\n";
+ return;
+ }
+ SarifOS << llvm::formatv("{0:2}", Document);
+ }
+
void reportFix(const DiagnosticBuilder &Diag,
const llvm::StringMap<Replacements> &Fix) {
for (const auto &FileAndReplacements : Fix) {
@@ -328,6 +449,11 @@ class ErrorReporter {
unsigned TotalFixes = 0U;
unsigned AppliedFixes = 0U;
unsigned WarningsAsErrors = 0U;
+ std::string SarifExportPath;
+ std::optional<SarifDocumentWriter> SarifWriter;
+ // SarifRuleIdx is used across multiple calls of exportSarifResult as cache to
+ // avoid creating duplicate SARIF rules.
+ llvm::StringMap<size_t> SarifRuleIdx;
};
class ClangTidyASTConsumer : public MultiplexConsumer {
@@ -613,7 +739,8 @@ runClangTidy(ClangTidyContext &Context, const CompilationDatabase &Compilations,
Context.setEnableProfiling(EnableCheckProfile);
Context.setProfileStoragePrefix(StoreCheckProfile);
- ClangTidyDiagnosticConsumer DiagConsumer(Context, nullptr, true, ApplyAnyFix);
+ ClangTidyDiagnosticConsumer DiagConsumer(Context, nullptr, true, ApplyAnyFix,
+ true);
auto DiagOpts = std::make_unique<DiagnosticOptions>();
DiagnosticsEngine DE(DiagnosticIDs::create(), *DiagOpts, &DiagConsumer,
/*ShouldOwnClient=*/false);
@@ -668,8 +795,9 @@ runClangTidy(ClangTidyContext &Context, const CompilationDatabase &Compilations,
void handleErrors(llvm::ArrayRef<ClangTidyError> Errors,
ClangTidyContext &Context, FixBehaviour Fix,
unsigned &WarningsAsErrorsCount,
- llvm::IntrusiveRefCntPtr<llvm::vfs::FileSystem> BaseFS) {
- ErrorReporter Reporter(Context, Fix, std::move(BaseFS));
+ llvm::IntrusiveRefCntPtr<llvm::vfs::FileSystem> BaseFS,
+ StringRef SarifExportPath) {
+ ErrorReporter Reporter(Context, Fix, std::move(BaseFS), SarifExportPath);
llvm::vfs::FileSystem &FileSystem =
Reporter.getSourceManager().getFileManager().getVirtualFileSystem();
auto InitialWorkingDir = FileSystem.getCurrentWorkingDirectory();
diff --git a/clang-tools-extra/clang-tidy/ClangTidy.h b/clang-tools-extra/clang-tidy/ClangTidy.h
index 5fac25bd3bbca3..ee750b2339e626 100644
--- a/clang-tools-extra/clang-tidy/ClangTidy.h
+++ b/clang-tools-extra/clang-tidy/ClangTidy.h
@@ -118,7 +118,8 @@ enum FixBehaviour {
void handleErrors(llvm::ArrayRef<ClangTidyError> Errors,
ClangTidyContext &Context, FixBehaviour Fix,
unsigned &WarningsAsErrorsCount,
- llvm::IntrusiveRefCntPtr<llvm::vfs::FileSystem> BaseFS);
+ llvm::IntrusiveRefCntPtr<llvm::vfs::FileSystem> BaseFS,
+ StringRef SarifExportPath = {});
/// Serializes replacements into YAML and writes them to the specified
/// output stream.
diff --git a/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.cpp b/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.cpp
index cd163300ec896f..6af3535de85fbb 100644
--- a/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.cpp
+++ b/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.cpp
@@ -70,37 +70,29 @@ class ClangTidyDiagnosticRenderer : public DiagnosticRenderer {
? tooling::DiagnosticMessage(Message, Loc.getManager(), Loc)
: tooling::DiagnosticMessage(Message);
- // Make sure that if a TokenRange is received from the check it is unfurled
- // into a real CharRange for the diagnostic printer later.
- // Whatever we store here gets decoupled from the current SourceManager, so
- // we **have to** know the exact position and length of the highlight.
- const auto ToCharRange = [this, &Loc](const CharSourceRange &SourceRange) {
- if (SourceRange.isCharRange())
- return SourceRange;
- assert(SourceRange.isTokenRange());
- const SourceLocation End = Lexer::getLocForEndOfToken(
- SourceRange.getEnd(), 0, Loc.getManager(), LangOpts);
- return CharSourceRange::getCharRange(SourceRange.getBegin(), End);
+ // Whatever we store here gets decoupled from the current Source Manager, so
+ // each range must become a character range in a real file now.
+ // makeFileCharRange maps a range covering a whole macro expansion to the
+ // macro call, and a range inside a macro argument to where the argument is
+ // written. Ranges it cannot map (invalid, or only part of a macro body)
+ // come back invalid and are dropped.
+ const auto StoreRanges = [&](tooling::DiagnosticMessage &Stored) {
+ for (const CharSourceRange &Range : Ranges) {
+ const CharSourceRange FileRange =
+ Lexer::makeFileCharRange(Range, Loc.getManager(), LangOpts);
+ if (FileRange.isValid())
+ Stored.Ranges.emplace_back(Loc.getManager(), FileRange);
+ }
};
- // We are only interested in valid ranges.
- const auto ValidRanges =
- llvm::make_filter_range(Ranges, [](const CharSourceRange &R) {
- return R.getAsRange().isValid();
- });
-
if (Level == DiagnosticsEngine::Note) {
Error.Notes.push_back(TidyMessage);
- for (const CharSourceRange &SourceRange : ValidRanges)
- Error.Notes.back().Ranges.emplace_back(Loc.getManager(),
- ToCharRange(SourceRange));
+ StoreRanges(Error.Notes.back());
return;
}
assert(Error.Message.Message.empty() && "Overwriting a diagnostic message");
Error.Message = TidyMessage;
- for (const CharSourceRange &SourceRange : ValidRanges)
- Error.Message.Ranges.emplace_back(Loc.getManager(),
- ToCharRange(SourceRange));
+ StoreRanges(Error.Message);
}
void emitDiagnosticLoc(FullSourceLoc Loc, PresumedLoc PLoc,
diff --git a/clang-tools-extra/clang-tidy/tool/ClangTidyMain.cpp b/clang-tools-extra/clang-tidy/tool/ClangTidyMain.cpp
index 7ee556fa7c51fa..726fbd9b6e30d7 100644
--- a/clang-tools-extra/clang-tidy/tool/ClangTidyMain.cpp
+++ b/clang-tools-extra/clang-tidy/tool/ClangTidyMain.cpp
@@ -21,6 +21,7 @@
#include "clang/Tooling/CommonOptionsParser.h"
#include "llvm/ADT/StringSet.h"
#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/FileSystem.h"
#include "llvm/Support/InitLLVM.h"
#include "llvm/Support/PluginLoader.h" // IWYU pragma: keep
#include "llvm/Support/Process.h"
@@ -311,6 +312,14 @@ code with clang-apply-replacements.
cl::value_desc("filename"),
cl::cat(ClangTidyCategory));
+static cl::opt<std::string> SarifExport("experimental-sarif-export", desc(R"(
+File in which to store diagnostics in SARIF format.
+Use '-' to write to stdout instead of the
+usual diagnostic output.
+)"),
+ cl::value_desc("filename"),
+ cl::cat(ClangTidyCategory));
+
static cl::opt<bool> Quiet("quiet", desc(R"(
Run clang-tidy in quiet mode. This suppresses
printing statistics about ignored warnings and
@@ -634,6 +643,15 @@ int clangTidyMain(int argc, const char **argv) {
return 1;
}
+ if (!SarifExport.empty() && SarifExport == ExportFixes &&
+ !llvm::sys::fs::equivalent(SarifExport, ExportFixes)) {
+ llvm::errs() << "Error: --experimental-sarif-export and --export-fixes "
+ "cannot write to "
+ "the same output '"
+ << SarifExport << "'.\n";
+ return 1;
+ }
+
const llvm::IntrusiveRefCntPtr<vfs::OverlayFileSystem> BaseFS =
createBaseFS();
if (!BaseFS)
@@ -730,6 +748,19 @@ int clangTidyMain(int argc, const char **argv) {
return 1;
}
+ // Checked after the informational modes so they never truncate the file.
+ if (!SarifExport.empty() && SarifExport != "-") {
+ std::error_code EC;
+ {
+ const llvm::raw_fd_ostream OS(SarifExport, EC, llvm::sys::fs::OF_Text);
+ }
+ if (EC) {
+ llvm::errs() << "Error opening SARIF export file '" << SarifExport
+ << "': " << EC.message() << '\n';
+ return 1;
+ }
+ }
+
llvm::InitializeAllTargetInfos();
llvm::InitializeAllTargetMCs();
llvm::InitializeAllAsmParsers();
@@ -754,7 +785,7 @@ int clangTidyMain(int argc, const char **argv) {
unsigned WErrorCount = 0;
handleErrors(Errors, Context, DisableFixes ? FB_NoFix : Behaviour,
- WErrorCount, BaseFS);
+ WErrorCount, BaseFS, SarifExport);
if (!ExportFixes.empty() && !Errors.empty()) {
std::error_code EC;
diff --git a/clang-tools-extra/docs/ReleaseNotes.md b/clang-tools-extra/docs/ReleaseNotes.md
index a8ed116d635d04..30a58ae13a511d 100644
--- a/clang-tools-extra/docs/ReleaseNotes.md
+++ b/clang-tools-extra/docs/ReleaseNotes.md
@@ -143,6 +143,10 @@ infrastructure are described first, followed by tool-specific sections.
`ImplementationFileExtensions` contains an invalid entry, such as one with a
leading dot. The invalid option is now reported as a warning.
+- Added a new {program}`clang-tidy` command line option `-sarif-export=<file>`
+ in which clang-tidy diagnostics are exported in the [SARIF](https://sarifweb.azurewebsites.net/)
+ format.
+
#### New checks
- New {doc}`bugprone-macro-condition
diff --git a/clang-tools-extra/docs/clang-tidy/index.rst b/clang-tools-extra/docs/clang-tidy/index.rst
index 9325307b6c9e49..cf5f68ee69506c 100644
--- a/clang-tools-extra/docs/clang-tidy/index.rst
+++ b/clang-tools-extra/docs/clang-tidy/index.rst
@@ -184,6 +184,9 @@ An overview of all the command-line options:
--experimental-custom-checks - Enable experimental clang-query based
custom checks.
see https://clang.llvm.org/extra/clang-tidy/QueryBasedCustomChecks.html.
+ --experimental-sarif-export=<filename> - File in which to store diagnostics in SARIF format.
+ Use '-' to write to stdout instead of the usual
+ diagnostic output.
--explain-config - For each enabled check explains, where it is
enabled, i.e. in clang-tidy binary, command
line or a specific configuration file.
diff --git a/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-conflicting-output.cpp b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-conflicting-output.cpp
new file mode 100644
index 00000000000000..f8ae706a5104ac
--- /dev/null
+++ b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-conflicting-output.cpp
@@ -0,0 +1,8 @@
+// RUN: not clang-tidy %s -checks='-*,google-explicit-constructor' --experimental-sarif-export=- -export-fixes=- -- 2>&1 | FileCheck %s --check-prefix=CHECK-STDOUT
+// RUN: not clang-tidy %s -checks='-*,google-explicit-constructor' --experimental-sarif-export=%t.out -export-fixes=%t.out -- 2>&1 | FileCheck %s --check-prefix=CHECK-FILE
+
+// CHECK-STDOUT: Error: --experimental-sarif-export and --export-fixes cannot write to the same output '-'.
+// CHECK-STDOUT-NOT: warning:
+// CHECK-FILE: Error: --experimental-sarif-export and --export-fixes cannot write to the same output '{{.*}}.out'.
+
+struct A { A(int) {} };
diff --git a/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-EOF.cpp b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-EOF.cpp
new file mode 100644
index 00000000000000..6d31ab6ec00f8b
--- /dev/null
+++ b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-EOF.cpp
@@ -0,0 +1,76 @@
+// RUN: echo -n 'int main() { return 0; }' > %t-input.cpp
+// RUN: clang-tidy %t-input.cpp -checks='-*,clang-diagnostic-newline-eof' --allow-no-checks --experimental-sarif-export=%t.sarif -- -Wnewline-eof > %t.msg 2>&1
+// RUN: FileCheck -input-file=%t.msg -check-prefix=CHECK-MESSAGES %s -implicit-check-not='{{warning|error|note}}:'
+// RUN: FileCheck -input-file=%t.sarif -check-prefix=CHECK-SARIF %s
+
+//CHECK-MESSAGES: -input.cpp:1:25: warning: no newline at end of file [clang-diagnostic-newline-eof]
+
+//CHECK-SARIF: {
+//CHECK-SARIF-NEXT: "$schema": "https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json",
+//CHECK-SARIF-NEXT: "runs": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "artifacts": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "length": {{[0-9]+}},
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "mimeType": "text/plain",
+//CHECK-SARIF-NEXT: "roles": [
+//CHECK-SARIF-NEXT: "resultFile"
+//CHECK-SARIF-NEXT: ]
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "columnKind": "unicodeCodePoints",
+//CHECK-SARIF-NEXT: "results": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 25,
+//CHECK-SARIF-NEXT: "startColumn": 25,
+//CHECK-SARIF-NEXT: "startLine": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "no newline at end of file"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-newline-eof",
+//CHECK-SARIF-NEXT: "ruleIndex": 0
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "tool": {
+//CHECK-SARIF-NEXT: "driver": {
+//CHECK-SARIF-NEXT: "fullName": "clang-tidy",
+//CHECK-SARIF-NEXT: "informationUri": "https://clang.llvm.org/extra/clang-tidy/",
+//CHECK-SARIF-NEXT: "language": "en-US",
+//CHECK-SARIF-NEXT: "name": "clang-tidy",
+//CHECK-SARIF-NEXT: "rules": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "rank": -1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "clang-diagnostic-newline-eof",
+//CHECK-SARIF-NEXT: "name": "clang-diagnostic-newline-eof"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
diff --git a/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-macro-thread-flows.cpp b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-macro-thread-flows.cpp
new file mode 100644
index 00000000000000..1db2806a8db568
--- /dev/null
+++ b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-macro-thread-flows.cpp
@@ -0,0 +1,175 @@
+// RUN: grep -Ev "// *[A-Z-]+:" %s > %t-input.cpp
+// RUN: clang-tidy %t-input.cpp -checks='-*,clang-analyzer-*' --experimental-sarif-export=%t.sarif > %t.msg 2>&1
+// RUN: FileCheck -input-file=%t.msg -check-prefix=CHECK-MESSAGES %s -implicit-check-not='{{warning|error|note}}:'
+// RUN: FileCheck -input-file=%t.sarif -check-prefix=CHECK-SARIF %s
+#define DEREF(p) (*(p))
+#define NUL nullptr
+int g() { int *ptr = NUL; return DEREF(ptr); }
+
+//CHECK-MESSAGES: -input.cpp:3:34: warning: Dereference of null pointer (loaded from variable 'ptr') [clang-analyzer-core.NullDereference]
+//CHECK-MESSAGES: -input.cpp:1:18: note: expanded from macro 'DEREF'
+//CHECK-MESSAGES: -input.cpp:3:11: note: 'ptr' initialized to a null pointer value
+//CHECK-MESSAGES: -input.cpp:3:34: note: Dereference of null pointer (loaded from variable 'ptr')
+//CHECK-MESSAGES: -input.cpp:1:18: note: expanded from macro 'DEREF'
+
+//CHECK-SARIF: {
+//CHECK-SARIF-NEXT: "$schema": "https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json",
+//CHECK-SARIF-NEXT: "runs": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "artifacts": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "length": {{[0-9]+}},
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "mimeType": "text/plain",
+//CHECK-SARIF-NEXT: "roles": [
+//CHECK-SARIF-NEXT: "resultFile"
+//CHECK-SARIF-NEXT: ]
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "columnKind": "unicodeCodePoints",
+//CHECK-SARIF-NEXT: "results": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "codeFlows": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "threadFlows": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "importance": "important",
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "expanded from macro 'DEREF'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 19,
+//CHECK-SARIF-NEXT: "endLine": 1,
+//CHECK-SARIF-NEXT: "startColumn": 18,
+//CHECK-SARIF-NEXT: "startLine": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "importance": "important",
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "'ptr' initialized to a null pointer value"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 12,
+//CHECK-SARIF-NEXT: "endLine": 3,
+//CHECK-SARIF-NEXT: "startColumn": 11,
+//CHECK-SARIF-NEXT: "startLine": 3
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "importance": "important",
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "Dereference of null pointer (loaded from variable 'ptr')"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 35,
+//CHECK-SARIF-NEXT: "endLine": 3,
+//CHECK-SARIF-NEXT: "startColumn": 34,
+//CHECK-SARIF-NEXT: "startLine": 3
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "importance": "important",
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "expanded from macro 'DEREF'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 19,
+//CHECK-SARIF-NEXT: "endLine": 1,
+//CHECK-SARIF-NEXT: "startColumn": 18,
+//CHECK-SARIF-NEXT: "startLine": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ]
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ]
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 35,
+//CHECK-SARIF-NEXT: "endLine": 3,
+//CHECK-SARIF-NEXT: "startColumn": 34,
+//CHECK-SARIF-NEXT: "startLine": 3
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "Dereference of null pointer (loaded from variable 'ptr')"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-analyzer-core.NullDereference",
+//CHECK-SARIF-NEXT: "ruleIndex": 0
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "tool": {
+//CHECK-SARIF-NEXT: "driver": {
+//CHECK-SARIF-NEXT: "fullName": "clang-tidy",
+//CHECK-SARIF-NEXT: "informationUri": "https://clang.llvm.org/extra/clang-tidy/",
+//CHECK-SARIF-NEXT: "language": "en-US",
+//CHECK-SARIF-NEXT: "name": "clang-tidy",
+//CHECK-SARIF-NEXT: "rules": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "rank": -1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "clang-analyzer-core.NullDereference",
+//CHECK-SARIF-NEXT: "name": "clang-analyzer-core.NullDereference"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
diff --git a/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-stdout.cpp b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-stdout.cpp
new file mode 100644
index 00000000000000..a873fd64c57aeb
--- /dev/null
+++ b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics-stdout.cpp
@@ -0,0 +1,302 @@
+// RUN: grep -Ev "// *[A-Z-]+:" %s > %t-input.cpp
+// RUN: not clang-tidy %t-input.cpp -checks='-*,google-explicit-constructor,clang-diagnostic-missing-prototypes,clang-diagnostic-zero-length-array' --warnings-as-errors='clang-diagnostic-missing-prototypes,google-explicit-constructor' --experimental-sarif-export=- -- -Wmissing-prototypes -Wzero-length-array 2>/dev/null | FileCheck %s
+#define X(n) void n ## n() {}
+X(f)
+int a[-1];
+int b[0];
+
+void test(x);
+struct Foo {
+ member;
+ Foo(int) {}
+};
+
+//CHECK: {{^{$}}
+//CHECK-NEXT: "$schema": "https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json",
+//CHECK-NEXT: "runs": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "artifacts": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "length": {{[0-9]+}},
+//CHECK-NEXT: "location": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "mimeType": "text/plain",
+//CHECK-NEXT: "roles": [
+//CHECK-NEXT: "resultFile"
+//CHECK-NEXT: ]
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "columnKind": "unicodeCodePoints",
+//CHECK-NEXT: "results": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "locations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 2,
+//CHECK-NEXT: "endLine": 2,
+//CHECK-NEXT: "startColumn": 1,
+//CHECK-NEXT: "startLine": 2
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "no previous prototype for function 'ff'"
+//CHECK-NEXT: },
+//CHECK-NEXT: "relatedLocations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "id": 0,
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "expanded from macro 'X'"
+//CHECK-NEXT: },
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 20,
+//CHECK-NEXT: "endLine": 1,
+//CHECK-NEXT: "startColumn": 19,
+//CHECK-NEXT: "startLine": 1
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "id": 1,
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "expanded from here"
+//CHECK-NEXT: }
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "id": 2,
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "declare 'static' if the function is not intended to be used outside of this translation unit"
+//CHECK-NEXT: },
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 2,
+//CHECK-NEXT: "endLine": 2,
+//CHECK-NEXT: "startColumn": 1,
+//CHECK-NEXT: "startLine": 2
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "id": 3,
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "expanded from macro 'X'"
+//CHECK-NEXT: },
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 15,
+//CHECK-NEXT: "endLine": 1,
+//CHECK-NEXT: "startColumn": 14,
+//CHECK-NEXT: "startLine": 1
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "ruleId": "clang-diagnostic-missing-prototypes",
+//CHECK-NEXT: "ruleIndex": 0
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "locations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 8,
+//CHECK-NEXT: "endLine": 3,
+//CHECK-NEXT: "startColumn": 7,
+//CHECK-NEXT: "startLine": 3
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "'a' declared as an array with a negative size"
+//CHECK-NEXT: },
+//CHECK-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-NEXT: "ruleIndex": 1
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "level": "warning",
+//CHECK-NEXT: "locations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 8,
+//CHECK-NEXT: "endLine": 4,
+//CHECK-NEXT: "startColumn": 7,
+//CHECK-NEXT: "startLine": 4
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "zero size arrays are an extension"
+//CHECK-NEXT: },
+//CHECK-NEXT: "ruleId": "clang-diagnostic-zero-length-array",
+//CHECK-NEXT: "ruleIndex": 2
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "locations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 12,
+//CHECK-NEXT: "endLine": 6,
+//CHECK-NEXT: "startColumn": 11,
+//CHECK-NEXT: "startLine": 6
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "unknown type name 'x'"
+//CHECK-NEXT: },
+//CHECK-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-NEXT: "ruleIndex": 1
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "locations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 4,
+//CHECK-NEXT: "endLine": 8,
+//CHECK-NEXT: "startColumn": 3,
+//CHECK-NEXT: "startLine": 8
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "a type specifier is required for all declarations"
+//CHECK-NEXT: },
+//CHECK-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-NEXT: "ruleIndex": 1
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "locations": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "physicalLocation": {
+//CHECK-NEXT: "artifactLocation": {
+//CHECK-NEXT: "index": 0,
+//CHECK-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-NEXT: },
+//CHECK-NEXT: "region": {
+//CHECK-NEXT: "endColumn": 4,
+//CHECK-NEXT: "endLine": 9,
+//CHECK-NEXT: "startColumn": 3,
+//CHECK-NEXT: "startLine": 9
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "message": {
+//CHECK-NEXT: "text": "single-argument constructors must be marked explicit to avoid unintentional implicit conversions"
+//CHECK-NEXT: },
+//CHECK-NEXT: "ruleId": "google-explicit-constructor",
+//CHECK-NEXT: "ruleIndex": 3
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "tool": {
+//CHECK-NEXT: "driver": {
+//CHECK-NEXT: "fullName": "clang-tidy",
+//CHECK-NEXT: "informationUri": "https://clang.llvm.org/extra/clang-tidy/",
+//CHECK-NEXT: "language": "en-US",
+//CHECK-NEXT: "name": "clang-tidy",
+//CHECK-NEXT: "rules": [
+//CHECK-NEXT: {
+//CHECK-NEXT: "defaultConfiguration": {
+//CHECK-NEXT: "enabled": true,
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "rank": 50
+//CHECK-NEXT: },
+//CHECK-NEXT: "fullDescription": {
+//CHECK-NEXT: "text": ""
+//CHECK-NEXT: },
+//CHECK-NEXT: "id": "clang-diagnostic-missing-prototypes",
+//CHECK-NEXT: "name": "clang-diagnostic-missing-prototypes"
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "defaultConfiguration": {
+//CHECK-NEXT: "enabled": true,
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "rank": 50
+//CHECK-NEXT: },
+//CHECK-NEXT: "fullDescription": {
+//CHECK-NEXT: "text": ""
+//CHECK-NEXT: },
+//CHECK-NEXT: "id": "clang-diagnostic-error",
+//CHECK-NEXT: "name": "clang-diagnostic-error"
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "defaultConfiguration": {
+//CHECK-NEXT: "enabled": true,
+//CHECK-NEXT: "level": "warning",
+//CHECK-NEXT: "rank": -1
+//CHECK-NEXT: },
+//CHECK-NEXT: "fullDescription": {
+//CHECK-NEXT: "text": ""
+//CHECK-NEXT: },
+//CHECK-NEXT: "id": "clang-diagnostic-zero-length-array",
+//CHECK-NEXT: "name": "clang-diagnostic-zero-length-array"
+//CHECK-NEXT: },
+//CHECK-NEXT: {
+//CHECK-NEXT: "defaultConfiguration": {
+//CHECK-NEXT: "enabled": true,
+//CHECK-NEXT: "level": "error",
+//CHECK-NEXT: "rank": 50
+//CHECK-NEXT: },
+//CHECK-NEXT: "fullDescription": {
+//CHECK-NEXT: "text": ""
+//CHECK-NEXT: },
+//CHECK-NEXT: "id": "google-explicit-constructor",
+//CHECK-NEXT: "name": "google-explicit-constructor"
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "version": "{{.*}}"
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: }
+//CHECK-NEXT: ],
+//CHECK-NEXT: "version": "{{.*}}"
+//CHECK-NEXT: }
diff --git a/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics.cpp b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics.cpp
new file mode 100644
index 00000000000000..44e5ef51671e56
--- /dev/null
+++ b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-diagnostics.cpp
@@ -0,0 +1,318 @@
+// RUN: grep -Ev "// *[A-Z-]+:" %s > %t-input.cpp
+// RUN: not clang-tidy %t-input.cpp -checks='-*,google-explicit-constructor,cppcoreguidelines-explicit-constructor,clang-diagnostic-missing-prototypes,clang-diagnostic-zero-length-array' --warnings-as-errors='clang-diagnostic-missing-prototypes,google-explicit-constructor' --experimental-sarif-export=%t.sarif -- -Wmissing-prototypes -Wzero-length-array > %t.msg 2>&1
+// RUN: FileCheck -input-file=%t.msg -check-prefix=CHECK-MESSAGES %s -implicit-check-not='{{warning|error|note}}:'
+// RUN: FileCheck -input-file=%t.sarif -check-prefix=CHECK-SARIF %s
+#define X(n) void n ## n() {}
+X(f)
+int a[-1];
+int b[0];
+
+void test(x);
+struct Foo {
+ member;
+ Foo(int) {}
+};
+
+//CHECK-MESSAGES: -input.cpp:2:1: error: no previous prototype for function 'ff' [clang-diagnostic-missing-prototypes,-warnings-as-errors]
+//CHECK-MESSAGES: -input.cpp:1:19: note: expanded from macro 'X'
+//CHECK-MESSAGES: {{^}}note: expanded from here{{$}}
+//CHECK-MESSAGES: -input.cpp:2:1: note: declare 'static' if the function is not intended to be used outside of this translation unit
+//CHECK-MESSAGES: -input.cpp:1:14: note: expanded from macro 'X'
+//CHECK-MESSAGES: -input.cpp:3:7: error: 'a' declared as an array with a negative size [clang-diagnostic-error]
+//CHECK-MESSAGES: -input.cpp:4:7: warning: zero size arrays are an extension [clang-diagnostic-zero-length-array]
+//CHECK-MESSAGES: -input.cpp:6:11: error: unknown type name 'x' [clang-diagnostic-error]
+//CHECK-MESSAGES: -input.cpp:8:3: error: a type specifier is required for all declarations [clang-diagnostic-error]
+//CHECK-MESSAGES: -input.cpp:9:3: error: single-argument constructors must be marked explicit to avoid unintentional implicit conversions [cppcoreguidelines-explicit-constructor,google-explicit-constructor,-warnings-as-errors]
+
+//CHECK-SARIF: {
+//CHECK-SARIF-NEXT: "$schema": "https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json",
+//CHECK-SARIF-NEXT: "runs": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "artifacts": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "length": {{[0-9]+}},
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "mimeType": "text/plain",
+//CHECK-SARIF-NEXT: "roles": [
+//CHECK-SARIF-NEXT: "resultFile"
+//CHECK-SARIF-NEXT: ]
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "columnKind": "unicodeCodePoints",
+//CHECK-SARIF-NEXT: "results": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 2,
+//CHECK-SARIF-NEXT: "endLine": 2,
+//CHECK-SARIF-NEXT: "startColumn": 1,
+//CHECK-SARIF-NEXT: "startLine": 2
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "no previous prototype for function 'ff'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "relatedLocations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "id": 0,
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "expanded from macro 'X'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 20,
+//CHECK-SARIF-NEXT: "endLine": 1,
+//CHECK-SARIF-NEXT: "startColumn": 19,
+//CHECK-SARIF-NEXT: "startLine": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "id": 1,
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "expanded from here"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "id": 2,
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "declare 'static' if the function is not intended to be used outside of this translation unit"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 2,
+//CHECK-SARIF-NEXT: "endLine": 2,
+//CHECK-SARIF-NEXT: "startColumn": 1,
+//CHECK-SARIF-NEXT: "startLine": 2
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "id": 3,
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "expanded from macro 'X'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 15,
+//CHECK-SARIF-NEXT: "endLine": 1,
+//CHECK-SARIF-NEXT: "startColumn": 14,
+//CHECK-SARIF-NEXT: "startLine": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-missing-prototypes",
+//CHECK-SARIF-NEXT: "ruleIndex": 0
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 8,
+//CHECK-SARIF-NEXT: "endLine": 3,
+//CHECK-SARIF-NEXT: "startColumn": 7,
+//CHECK-SARIF-NEXT: "startLine": 3
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "'a' declared as an array with a negative size"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-SARIF-NEXT: "ruleIndex": 1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 8,
+//CHECK-SARIF-NEXT: "endLine": 4,
+//CHECK-SARIF-NEXT: "startColumn": 7,
+//CHECK-SARIF-NEXT: "startLine": 4
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "zero size arrays are an extension"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-zero-length-array",
+//CHECK-SARIF-NEXT: "ruleIndex": 2
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 12,
+//CHECK-SARIF-NEXT: "endLine": 6,
+//CHECK-SARIF-NEXT: "startColumn": 11,
+//CHECK-SARIF-NEXT: "startLine": 6
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "unknown type name 'x'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-SARIF-NEXT: "ruleIndex": 1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 4,
+//CHECK-SARIF-NEXT: "endLine": 8,
+//CHECK-SARIF-NEXT: "startColumn": 3,
+//CHECK-SARIF-NEXT: "startLine": 8
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "a type specifier is required for all declarations"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-SARIF-NEXT: "ruleIndex": 1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 4,
+//CHECK-SARIF-NEXT: "endLine": 9,
+//CHECK-SARIF-NEXT: "startColumn": 3,
+//CHECK-SARIF-NEXT: "startLine": 9
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "single-argument constructors must be marked explicit to avoid unintentional implicit conversions"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "properties": {
+//CHECK-SARIF-NEXT: "enabledAliases": "google-explicit-constructor"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "cppcoreguidelines-explicit-constructor",
+//CHECK-SARIF-NEXT: "ruleIndex": 3
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "tool": {
+//CHECK-SARIF-NEXT: "driver": {
+//CHECK-SARIF-NEXT: "fullName": "clang-tidy",
+//CHECK-SARIF-NEXT: "informationUri": "https://clang.llvm.org/extra/clang-tidy/",
+//CHECK-SARIF-NEXT: "language": "en-US",
+//CHECK-SARIF-NEXT: "name": "clang-tidy",
+//CHECK-SARIF-NEXT: "rules": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "rank": 50
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "clang-diagnostic-missing-prototypes",
+//CHECK-SARIF-NEXT: "name": "clang-diagnostic-missing-prototypes"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "rank": 50
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "clang-diagnostic-error",
+//CHECK-SARIF-NEXT: "name": "clang-diagnostic-error"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "rank": -1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "clang-diagnostic-zero-length-array",
+//CHECK-SARIF-NEXT: "name": "clang-diagnostic-zero-length-array"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "rank": 50
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "cppcoreguidelines-explicit-constructor",
+//CHECK-SARIF-NEXT: "name": "cppcoreguidelines-explicit-constructor"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
diff --git a/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-no-location.cpp b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-no-location.cpp
new file mode 100644
index 00000000000000..8e9146c9f8a441
--- /dev/null
+++ b/clang-tools-extra/test/clang-tidy/infrastructure/export-SARIF-no-location.cpp
@@ -0,0 +1,104 @@
+// RUN: grep -Ev "// *[A-Z-]+:" %s > %t-input.cpp
+// RUN: not clang-tidy %t-input.cpp -checks='-*,clang-diagnostic-*,google-explicit-constructor' --experimental-sarif-export=%t.sarif -- -fake-command > %t.msg 2>&1
+// RUN: FileCheck -input-file=%t.msg -check-prefix=CHECK-MESSAGES %s -implicit-check-not='{{warning|error|note}}:'
+// RUN: FileCheck -input-file=%t.sarif -check-prefix=CHECK-SARIF %s
+class A { A(int) {} };
+
+//NOTE: "-fake-command" is rejected by the driver before a SourceManager is created,
+//NOTE: so the diagnostic is stored with an empty FilePath and ranges. Therefore, getResultRanges()
+//NOTE: omits locations entirely.
+
+//CHECK-MESSAGES: error: unknown argument: '-fake-command' [clang-diagnostic-error]
+//CHECK-MESSAGES: -input.cpp:1:11: warning: single-argument constructors must be marked explicit to avoid unintentional implicit conversions [google-explicit-constructor]
+
+//CHECK-SARIF: {
+//CHECK-SARIF-NEXT: "$schema": "https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json",
+//CHECK-SARIF-NEXT: "runs": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "artifacts": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "length": {{[0-9]+}},
+//CHECK-SARIF-NEXT: "location": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "mimeType": "text/plain",
+//CHECK-SARIF-NEXT: "roles": [
+//CHECK-SARIF-NEXT: "resultFile"
+//CHECK-SARIF-NEXT: ]
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "columnKind": "unicodeCodePoints",
+//CHECK-SARIF-NEXT: "results": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "unknown argument: '-fake-command'"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "clang-diagnostic-error",
+//CHECK-SARIF-NEXT: "ruleIndex": 0
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "locations": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "physicalLocation": {
+//CHECK-SARIF-NEXT: "artifactLocation": {
+//CHECK-SARIF-NEXT: "index": 0,
+//CHECK-SARIF-NEXT: "uri": "file://{{.*}}-input.cpp"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "region": {
+//CHECK-SARIF-NEXT: "endColumn": 12,
+//CHECK-SARIF-NEXT: "endLine": 1,
+//CHECK-SARIF-NEXT: "startColumn": 11,
+//CHECK-SARIF-NEXT: "startLine": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "message": {
+//CHECK-SARIF-NEXT: "text": "single-argument constructors must be marked explicit to avoid unintentional implicit conversions"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "ruleId": "google-explicit-constructor",
+//CHECK-SARIF-NEXT: "ruleIndex": 1
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "tool": {
+//CHECK-SARIF-NEXT: "driver": {
+//CHECK-SARIF-NEXT: "fullName": "clang-tidy",
+//CHECK-SARIF-NEXT: "informationUri": "https://clang.llvm.org/extra/clang-tidy/",
+//CHECK-SARIF-NEXT: "language": "en-US",
+//CHECK-SARIF-NEXT: "name": "clang-tidy",
+//CHECK-SARIF-NEXT: "rules": [
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "error",
+//CHECK-SARIF-NEXT: "rank": 50
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "clang-diagnostic-error",
+//CHECK-SARIF-NEXT: "name": "clang-diagnostic-error"
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: {
+//CHECK-SARIF-NEXT: "defaultConfiguration": {
+//CHECK-SARIF-NEXT: "enabled": true,
+//CHECK-SARIF-NEXT: "level": "warning",
+//CHECK-SARIF-NEXT: "rank": -1
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "fullDescription": {
+//CHECK-SARIF-NEXT: "text": ""
+//CHECK-SARIF-NEXT: },
+//CHECK-SARIF-NEXT: "id": "google-explicit-constructor",
+//CHECK-SARIF-NEXT: "name": "google-explicit-constructor"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: }
+//CHECK-SARIF-NEXT: ],
+//CHECK-SARIF-NEXT: "version": "{{.*}}"
+//CHECK-SARIF-NEXT: }
diff --git a/clang/include/clang/Basic/Sarif.h b/clang/include/clang/Basic/Sarif.h
index c7afa2813f570c..4e5aa2fa6ab74d 100644
--- a/clang/include/clang/Basic/Sarif.h
+++ b/clang/include/clang/Basic/Sarif.h
@@ -332,9 +332,10 @@ class SarifResult {
std::string HostedViewerURI;
llvm::SmallDenseMap<StringRef, std::string, 4> PartialFingerprints;
llvm::SmallVector<CharSourceRange, 8> Locations;
- llvm::SmallVector<CharSourceRange, 8> RelatedLocations;
+ llvm::SmallVector<std::pair<CharSourceRange, std::string>> RelatedLocations;
llvm::SmallVector<ThreadFlow, 8> ThreadFlows;
std::optional<SarifResultLevel> LevelOverride;
+ llvm::StringMap<std::string> Properties;
SarifResult() = delete;
explicit SarifResult(uint32_t RuleIdx) : RuleIdx(RuleIdx) {}
@@ -373,7 +374,8 @@ class SarifResult {
return *this;
}
- SarifResult addRelatedLocations(llvm::ArrayRef<CharSourceRange> DiagLocs) {
+ SarifResult addRelatedLocations(llvm::ArrayRef<CharSourceRange> DiagLocs,
+ llvm::StringRef Message = "") {
#ifndef NDEBUG
for (const auto &Loc : DiagLocs) {
assert(
@@ -381,7 +383,8 @@ class SarifResult {
"SARIF RelatedLocations require character granular source ranges!");
}
#endif
- RelatedLocations.append(DiagLocs.begin(), DiagLocs.end());
+ for (const CharSourceRange &Loc : DiagLocs)
+ RelatedLocations.emplace_back(Loc, Message.str());
return *this;
}
@@ -395,6 +398,11 @@ class SarifResult {
return *this;
}
+ SarifResult addProperty(llvm::StringRef Key, llvm::StringRef Value) {
+ Properties[Key] = Value.str();
+ return *this;
+ }
+
SarifResult addPartialFingerprint(llvm::StringRef key,
llvm::StringRef value) {
PartialFingerprints[key] = value;
@@ -447,8 +455,10 @@ class SarifDocumentWriter {
createThreadFlows(const llvm::ArrayRef<ThreadFlow> ThreadFlows);
/// Add the given \ref CharSourceRange to the SARIF document as a physical
- /// location, with its corresponding artifact.
- llvm::json::Object createPhysicalLocation(const CharSourceRange &R);
+ /// location, with its corresponding artifact. Returns std::nullopt if \p R
+ /// is invalid or does not map to a file.
+ std::optional<llvm::json::Object>
+ createPhysicalLocation(const CharSourceRange &R);
public:
SarifDocumentWriter() = delete;
@@ -463,7 +473,9 @@ class SarifDocumentWriter {
/// Each run requires specifying the tool that is generating reporting items.
void createRun(const llvm::StringRef ShortToolName,
const llvm::StringRef LongToolName,
- const llvm::StringRef ToolVersion = CLANG_VERSION_STRING);
+ const llvm::StringRef ToolVersion = CLANG_VERSION_STRING,
+ const llvm::StringRef InformationUri =
+ "https://clang.llvm.org/docs/UsersManual.html");
/// If there is a current run, end it.
///
diff --git a/clang/include/clang/Frontend/SARIFDiagnostic.h b/clang/include/clang/Frontend/SARIFDiagnostic.h
index 7a6f27eb3b9fab..cef565db1686a5 100644
--- a/clang/include/clang/Frontend/SARIFDiagnostic.h
+++ b/clang/include/clang/Frontend/SARIFDiagnostic.h
@@ -6,8 +6,9 @@
//
//===----------------------------------------------------------------------===//
//
-// This is a utility class that provides support for constructing a SARIF object
-// containing diagnostics.
+// This file implements SARIF diagnostic emission, including the SARIFDiagnostic
+// renderer and helper utilities for mapping diagnostic levels and annotating
+// SARIF rules.
//
//===----------------------------------------------------------------------===//
@@ -20,6 +21,14 @@
namespace clang {
+// Maps clang DiagnosticsEngine level to SARIF result levels.
+SarifResultLevel getSarifResultLevel(DiagnosticsEngine::Level level);
+
+// Creates configuration for a SARIF rule based on the diagnostic level and sets
+// the rank.
+SarifRule addDiagnosticLevelToRule(SarifRule Rule,
+ DiagnosticsEngine::Level Level);
+
class SARIFDiagnostic : public DiagnosticRenderer {
public:
SARIFDiagnostic(raw_ostream &OS, const LangOptions &LangOpts,
@@ -70,9 +79,6 @@ class SARIFDiagnostic : public DiagnosticRenderer {
getSarifLocation(FullSourceLoc Loc, PresumedLoc PLoc,
ArrayRef<CharSourceRange> Ranges);
- SarifRule addDiagnosticLevelToRule(SarifRule Rule,
- DiagnosticsEngine::Level Level);
-
llvm::StringRef emitFilename(StringRef Filename, const SourceManager &SM);
llvm::SmallVector<std::pair<FullSourceLoc, PresumedLoc>>
diff --git a/clang/lib/Basic/Sarif.cpp b/clang/lib/Basic/Sarif.cpp
index e829b9bb47ad94..dd95b0efa28aea 100644
--- a/clang/lib/Basic/Sarif.cpp
+++ b/clang/lib/Basic/Sarif.cpp
@@ -162,11 +162,16 @@ static json::Object createTextRegion(const SourceManager &SM,
return Region;
}
-static json::Object createLocation(json::Object &&PhysicalLocation,
- StringRef Message = "") {
- json::Object Ret{{"physicalLocation", std::move(PhysicalLocation)}};
+/// \internal
+/// SARIF allows a location with only a message and no physicalLocation.
+static json::Object
+createLocation(std::optional<json::Object> &&PhysicalLocation,
+ StringRef Message = "") {
+ json::Object Ret;
+ if (PhysicalLocation)
+ Ret["physicalLocation"] = std::move(*PhysicalLocation);
if (!Message.empty())
- Ret.insert({"message", createMessage(Message)});
+ Ret["message"] = createMessage(Message);
return Ret;
}
@@ -205,15 +210,16 @@ createThreadFlowLocation(json::Object &&Location,
}
/// @}
-json::Object
+std::optional<json::Object>
SarifDocumentWriter::createPhysicalLocation(const CharSourceRange &R) {
- assert(R.isValid() &&
- "Cannot create a physicalLocation from invalid SourceRange!");
+ if (R.isInvalid())
+ return std::nullopt;
assert(R.isCharRange() &&
"Cannot create a physicalLocation from a token range!");
FullSourceLoc Start{R.getBegin(), SourceMgr};
OptionalFileEntryRef FE = Start.getExpansionLoc().getFileEntryRef();
- assert(FE && "Diagnostic does not exist within a valid file!");
+ if (!FE)
+ return std::nullopt;
const std::string &FileURI = fileNameToURI(getFileName(*FE));
auto I = CurrentArtifacts.find(FileURI);
@@ -297,7 +303,11 @@ void SarifDocumentWriter::endRun() {
SmallVector<std::pair<StringRef, SarifArtifact>, 0> Vec;
for (const auto &[K, V] : CurrentArtifacts)
Vec.emplace_back(K, V);
- llvm::sort(Vec, llvm::less_first());
+ // artifactLocation.index refers to the position in this array.
+ llvm::sort(Vec, [](const std::pair<StringRef, SarifArtifact> &L,
+ const std::pair<StringRef, SarifArtifact> &R) {
+ return L.second.Location.Index < R.second.Location.Index;
+ });
for (const auto &[_, A] : Vec) {
json::Object Loc{{"uri", A.Location.URI}};
if (A.Location.Index.has_value()) {
@@ -328,8 +338,8 @@ SarifDocumentWriter::createThreadFlows(ArrayRef<ThreadFlow> ThreadFlows) {
json::Object Ret{{"locations", json::Array{}}};
json::Array Locs;
for (const auto &ThreadFlow : ThreadFlows) {
- json::Object PLoc = createPhysicalLocation(ThreadFlow.Range);
- json::Object Loc = createLocation(std::move(PLoc), ThreadFlow.Message);
+ json::Object Loc = createLocation(createPhysicalLocation(ThreadFlow.Range),
+ ThreadFlow.Message);
Locs.emplace_back(
createThreadFlowLocation(std::move(Loc), ThreadFlow.Importance));
}
@@ -344,21 +354,20 @@ SarifDocumentWriter::createCodeFlow(ArrayRef<ThreadFlow> ThreadFlows) {
void SarifDocumentWriter::createRun(StringRef ShortToolName,
StringRef LongToolName,
- StringRef ToolVersion) {
+ StringRef ToolVersion,
+ StringRef InformationUri) {
// Clear resources associated with a previous run.
endRun();
// Signify a new run has begun.
Closed = false;
- json::Object Tool{
- {"driver",
- json::Object{{"name", ShortToolName},
- {"fullName", LongToolName},
- {"language", "en-US"},
- {"version", ToolVersion},
- {"informationUri",
- "https://clang.llvm.org/docs/UsersManual.html"}}}};
+ json::Object Tool{{"driver", json::Object{{"name", ShortToolName},
+ {"fullName", LongToolName},
+ {"language", "en-US"},
+ {"version", ToolVersion},
+ {"informationUri", InformationUri},
+ {"rules", json::Array{}}}}};
json::Object TheRun{{"tool", std::move(Tool)},
{"results", {}},
{"artifacts", {}},
@@ -399,21 +408,34 @@ void SarifDocumentWriter::appendResult(const SarifResult &Result) {
Ret["hostedViewerUri"] = Result.HostedViewerURI;
}
- if (!Result.Locations.empty()) {
- json::Array Locs;
- for (auto &Range : Result.Locations) {
- Locs.emplace_back(createLocation(createPhysicalLocation(Range)));
- }
+ if (!Result.Properties.empty()) {
+ json::Object Properties;
+ for (const auto &[Key, Value] : Result.Properties)
+ Properties[Key.str()] = Value;
+ Ret["properties"] = std::move(Properties);
+ }
+
+ json::Array Locs;
+ for (const auto &Range : Result.Locations)
+ if (std::optional<json::Object> PL = createPhysicalLocation(Range))
+ Locs.emplace_back(createLocation(std::move(PL)));
+ if (!Locs.empty())
Ret["locations"] = std::move(Locs);
+
+ // relatedLocations must hold unique items, a distinct Id per location is
+ // added to keep notes unique.
+ json::Array ReLocs;
+ int64_t RelatedLocationId = 0;
+ for (const auto &[Range, Message] : Result.RelatedLocations) {
+ json::Object Loc = createLocation(createPhysicalLocation(Range), Message);
+ if (Loc.empty())
+ continue;
+ Loc["id"] = RelatedLocationId++;
+ ReLocs.emplace_back(std::move(Loc));
}
- if (!Result.RelatedLocations.empty()) {
- json::Array ReLocs;
- for (auto &Range : Result.RelatedLocations) {
- ReLocs.emplace_back(createLocation(createPhysicalLocation(Range)));
- }
+ if (!ReLocs.empty())
Ret["relatedLocations"] = std::move(ReLocs);
- }
if (!Result.PartialFingerprints.empty()) {
json::Object fingerprints = {};
diff --git a/clang/lib/Frontend/SARIFDiagnostic.cpp b/clang/lib/Frontend/SARIFDiagnostic.cpp
index f6f3feb7bd86cd..b6c23c554f075b 100644
--- a/clang/lib/Frontend/SARIFDiagnostic.cpp
+++ b/clang/lib/Frontend/SARIFDiagnostic.cpp
@@ -30,6 +30,44 @@
namespace clang {
+// SARIF reportingConfiguration.rank is given to each severity. Notes, remarks
+// and warnings keep the default -1.
+namespace rank {
+constexpr float Error = 50.0f;
+constexpr float Fatal = 100.0f;
+} // namespace rank
+
+SarifResultLevel getSarifResultLevel(DiagnosticsEngine::Level level) {
+ switch (level) {
+ case DiagnosticsEngine::Ignored:
+ llvm_unreachable("Invalid diagnostic type");
+ case DiagnosticsEngine::Note:
+ return SarifResultLevel::Note;
+ case DiagnosticsEngine::Remark:
+ return SarifResultLevel::None;
+ case DiagnosticsEngine::Warning:
+ return SarifResultLevel::Warning;
+ case DiagnosticsEngine::Error:
+ case DiagnosticsEngine::Fatal:
+ return SarifResultLevel::Error;
+ }
+ llvm_unreachable("Un-handled DiagnosticsEngine::Level occured.");
+}
+
+SarifRule addDiagnosticLevelToRule(SarifRule Rule,
+ DiagnosticsEngine::Level Level) {
+ SarifReportingConfiguration Config =
+ SarifReportingConfiguration::create().setLevel(
+ getSarifResultLevel(Level));
+
+ if (Level == DiagnosticsEngine::Error)
+ Config = Config.setRank(rank::Error);
+ else if (Level == DiagnosticsEngine::Fatal)
+ Config = Config.setRank(rank::Fatal);
+
+ return Rule.setDefaultConfiguration(Config);
+}
+
SARIFDiagnostic::SARIFDiagnostic(raw_ostream &OS, const LangOptions &LangOpts,
DiagnosticOptions &DiagOpts,
SarifDocumentWriter *Writer)
@@ -164,34 +202,6 @@ SARIFDiagnostic::getSarifLocation(FullSourceLoc Loc, PresumedLoc PLoc,
return Locations;
}
-SarifRule
-SARIFDiagnostic::addDiagnosticLevelToRule(SarifRule Rule,
- DiagnosticsEngine::Level Level) {
- auto Config = SarifReportingConfiguration::create();
-
- switch (Level) {
- case DiagnosticsEngine::Note:
- Config = Config.setLevel(SarifResultLevel::Note);
- break;
- case DiagnosticsEngine::Remark:
- Config = Config.setLevel(SarifResultLevel::None);
- break;
- case DiagnosticsEngine::Warning:
- Config = Config.setLevel(SarifResultLevel::Warning);
- break;
- case DiagnosticsEngine::Error:
- Config = Config.setLevel(SarifResultLevel::Error).setRank(50);
- break;
- case DiagnosticsEngine::Fatal:
- Config = Config.setLevel(SarifResultLevel::Error).setRank(100);
- break;
- case DiagnosticsEngine::Ignored:
- assert(false && "Invalid diagnostic type");
- }
-
- return Rule.setDefaultConfiguration(Config);
-}
-
llvm::StringRef SARIFDiagnostic::emitFilename(StringRef Filename,
const SourceManager &SM) {
if (DiagOpts.AbsolutePath) {
diff --git a/clang/unittests/Basic/SarifTest.cpp b/clang/unittests/Basic/SarifTest.cpp
index 42e85085d646eb..a9ee0119cbfe53 100644
--- a/clang/unittests/Basic/SarifTest.cpp
+++ b/clang/unittests/Basic/SarifTest.cpp
@@ -290,7 +290,7 @@ TEST_F(SarifDocumentWriterTest, checkSerializingResultsWithCustomRuleConfig) {
TEST_F(SarifDocumentWriterTest, checkSerializingArtifacts) {
// GIVEN:
const std::string ExpectedOutput =
- R"({"$schema":"https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json","runs":[{"artifacts":[{"length":40,"location":{"index":0,"uri":"file:///main.cpp"},"mimeType":"text/plain","roles":["resultFile"]}],"columnKind":"unicodeCodePoints","results":[{"level":"error","locations":[{"physicalLocation":{"artifactLocation":{"index":0,"uri":"file:///main.cpp"},"region":{"endColumn":14,"startColumn":14,"startLine":3}}}],"message":{"text":"expected ';' after top level declarator"},"relatedLocations":[{"physicalLocation":{"artifactLocation":{"index":0,"uri":"file:///main.cpp"},"region":{"endColumn":14,"startColumn":14,"startLine":3}}}],"ruleId":"clang.unittest","ruleIndex":0}],"tool":{"driver":{"fullName":"sarif test runner","informationUri":"https://clang.llvm.org/docs/UsersManual.html","language":"en-US","name":"sarif test","rules":[{"defaultConfiguration":{"enabled":true,"level":"warning","rank":-1},"fullDescription":{"text":"Example rule created during unit tests"},"id":"clang.unittest","name":"clang unit test"}],"version":"1.0.0"}}}],"version":"2.1.0"})";
+ R"({"$schema":"https://docs.oasis-open.org/sarif/sarif/v2.1.0/cos02/schemas/sarif-schema-2.1.0.json","runs":[{"artifacts":[{"length":40,"location":{"index":0,"uri":"file:///main.cpp"},"mimeType":"text/plain","roles":["resultFile"]}],"columnKind":"unicodeCodePoints","results":[{"level":"error","locations":[{"physicalLocation":{"artifactLocation":{"index":0,"uri":"file:///main.cpp"},"region":{"endColumn":14,"startColumn":14,"startLine":3}}}],"message":{"text":"expected ';' after top level declarator"},"relatedLocations":[{"id":0,"physicalLocation":{"artifactLocation":{"index":0,"uri":"file:///main.cpp"},"region":{"endColumn":14,"startColumn":14,"startLine":3}}}],"ruleId":"clang.unittest","ruleIndex":0}],"tool":{"driver":{"fullName":"sarif test runner","informationUri":"https://clang.llvm.org/docs/UsersManual.html","language":"en-US","name":"sarif test","rules":[{"defaultConfiguration":{"enabled":true,"level":"warning","rank":-1},"fullDescription":{"text":"Example rule created during unit tests"},"id":"clang.unittest","name":"clang unit test"}],"version":"1.0.0"}}}],"version":"2.1.0"})";
SarifDocumentWriter Writer{SourceMgr};
const SarifRule &Rule =
@@ -388,4 +388,21 @@ TEST_F(SarifDocumentWriterTest, checkSerializingCodeflows) {
ASSERT_THAT(Output, ::testing::StrEq(ExpectedOutput));
}
+TEST_F(SarifDocumentWriterTest, artifactIndexMatchesArrayPosition) {
+ SarifDocumentWriter Writer{SourceMgr};
+ Writer.createRun("sarif test", "sarif test runner");
+ unsigned RuleIdx = Writer.createRule(SarifRule::create());
+
+ // "/z.cpp" is seen first (index 0) but sorts after "/a.h" by name.
+ FileID ZFile = registerSource("/z.cpp", "int z;\n");
+ FileID AFile = registerSource("/a.h", "int a;\n");
+ Writer.appendResult(SarifResult::create(RuleIdx).addLocations(
+ {getFakeCharSourceRange(ZFile, {1, 1}, {1, 2}),
+ getFakeCharSourceRange(AFile, {1, 1}, {1, 2})}));
+
+ std::string Output = serializeSarifDocument(Writer.createDocument());
+ EXPECT_THAT(Output, ::testing::HasSubstr(
+ R"("location":{"index":0,"uri":"file:///z.cpp"})"));
+}
+
} // namespace
>From a7bacdae9a9b3eba7d5d97974714855cd6e9b556 Mon Sep 17 00:00:00 2001
From: Brad Smith <brad at comstyle.com>
Date: Sat, 21 Nov 2026 05:53:43 -0500
Subject: [PATCH 08/35] [Mips] Don't use trunc.w.fmt for fptosi on MIPS I
(#244859)
MIPS I has no trunc.w.fmt. fptosi from float selected trunc.w.s anyway,
since its pattern was marked ISA_MIPS1, and fptosi from double failed
to select.
On MIPS I use cvt.w.fmt with the rounding mode set to round toward
zero, then restore the FCSR. This is the same sequence the assembler
uses for the trunc.w.fmt macro.
cfc1/ctc1 have the same delay as mfc1/mtc1 before MIPS IV, so add them
to the FPU delay slot hazard list. This puts a nop after cfc1 before its
result is used and after ctc1 before the next FPU instruction.
Co-authored-by: Claude Opus 5.5 <noreply at anthropic.com>
---
llvm/lib/Target/Mips/MipsInstrFPU.td | 17 +++-
llvm/lib/Target/Mips/MipsInstrInfo.cpp | 2 +
llvm/lib/Target/Mips/MipsInstrInfo.td | 5 ++
llvm/lib/Target/Mips/MipsSEISelLowering.cpp | 38 +++++++++
llvm/lib/Target/Mips/MipsSEISelLowering.h | 2 +
llvm/test/CodeGen/Mips/mips1-fptosi.ll | 91 +++++++++++++++++++++
6 files changed, 154 insertions(+), 1 deletion(-)
create mode 100644 llvm/test/CodeGen/Mips/mips1-fptosi.ll
diff --git a/llvm/lib/Target/Mips/MipsInstrFPU.td b/llvm/lib/Target/Mips/MipsInstrFPU.td
index df8c334df6c9de..5d012d4e352cfe 100644
--- a/llvm/lib/Target/Mips/MipsInstrFPU.td
+++ b/llvm/lib/Target/Mips/MipsInstrFPU.td
@@ -913,6 +913,16 @@ def LoadImmDoubleFGR : MipsAsmPseudoInst<(outs StrictlyFGR64Opnd:$rd),
"li.d\t$rd, $fpimm">,
FGR_64, HARDFLOAT;
+// MIPS I has no trunc.w.fmt. These pseudos are expanded by a custom inserter
+// into cvt.w.fmt with the rounding mode set to round toward zero.
+let usesCustomInserter = 1, hasNoSchedulingInfo = 1 in {
+ def PseudoTRUNC_W_S_M1 : PseudoSE<(outs FGR32Opnd:$fd),
+ (ins FGR32Opnd:$fs), []>, HARDFLOAT;
+ def PseudoTRUNC_W_D32_M1 : PseudoSE<(outs FGR32Opnd:$fd),
+ (ins AFGR64Opnd:$fs), []>, FGR_32,
+ HARDFLOAT;
+}
+
def SDC1_M1 : MipsAsmPseudoInst<(outs AFGR64Opnd:$fd),
(ins MemBase:$base, mem_simm16:$offset),
"s.d\t$fd, ${offset}(${base})">,
@@ -1046,7 +1056,9 @@ let Predicates = [IsR5900] in {
(CVT_W_S FGR32Opnd:$src)>, ISA_MIPS1;
}
def : MipsPat<(MipsTruncIntFP FGR32Opnd:$src),
- (TRUNC_W_S FGR32Opnd:$src)>, ISA_MIPS1;
+ (TRUNC_W_S FGR32Opnd:$src)>, ISA_MIPS2;
+def : MipsPat<(MipsTruncIntFP FGR32Opnd:$src),
+ (PseudoTRUNC_W_S_M1 FGR32Opnd:$src)>, ISA_MIPS1_NOT_MIPS2;
def : MipsPat<(MipsMTC1_D64 GPR32Opnd:$src),
(MTC1_D64 GPR32Opnd:$src)>, ISA_MIPS1, FGR_64;
@@ -1064,6 +1076,9 @@ def : MipsPat<(f64 (any_sint_to_fp GPR32Opnd:$src)),
let AdditionalPredicates = [NotInMicroMips] in {
def : MipsPat<(MipsTruncIntFP AFGR64Opnd:$src),
(TRUNC_W_D32 AFGR64Opnd:$src)>, ISA_MIPS2, FGR_32;
+ def : MipsPat<(MipsTruncIntFP AFGR64Opnd:$src),
+ (PseudoTRUNC_W_D32_M1 AFGR64Opnd:$src)>, ISA_MIPS1_NOT_MIPS2,
+ FGR_32;
def : MipsPat<(f32 (any_fpround AFGR64Opnd:$src)),
(CVT_S_D32 AFGR64Opnd:$src)>, ISA_MIPS1, FGR_32;
def : MipsPat<(f64 (any_fpextend FGR32Opnd:$src)),
diff --git a/llvm/lib/Target/Mips/MipsInstrInfo.cpp b/llvm/lib/Target/Mips/MipsInstrInfo.cpp
index 6ec5840490bd0d..12ce81c80b3937 100644
--- a/llvm/lib/Target/Mips/MipsInstrInfo.cpp
+++ b/llvm/lib/Target/Mips/MipsInstrInfo.cpp
@@ -665,6 +665,8 @@ bool MipsInstrInfo::HasFPUDelaySlot(const MachineInstr &MI) const {
switch (MI.getOpcode()) {
case Mips::MTC1:
case Mips::MFC1:
+ case Mips::CTC1:
+ case Mips::CFC1:
case Mips::MTC1_D64:
case Mips::MFC1_D64:
case Mips::DMTC1:
diff --git a/llvm/lib/Target/Mips/MipsInstrInfo.td b/llvm/lib/Target/Mips/MipsInstrInfo.td
index d20cdd92b518ab..ba342022b1626c 100644
--- a/llvm/lib/Target/Mips/MipsInstrInfo.td
+++ b/llvm/lib/Target/Mips/MipsInstrInfo.td
@@ -172,6 +172,7 @@ def MipsSDR : SDNode<"MipsISD::SDR", SDTStore,
def HasMips2 : Predicate<"Subtarget->hasMips2()">,
AssemblerPredicate<(all_of FeatureMips2)>;
+def NotMips2 : Predicate<"!Subtarget->hasMips2()">;
def HasMips3_32 : Predicate<"Subtarget->hasMips3_32()">,
AssemblerPredicate<(all_of FeatureMips3_32)>;
def HasMips3_32r2 : Predicate<"Subtarget->hasMips3_32r2()">,
@@ -303,6 +304,10 @@ class SYM_64 { list<Predicate> SYMPredicates = [IsSym64]; }
class ISA_MIPS1 {
list<Predicate> EncodingPredicates = [HasStdEnc];
}
+class ISA_MIPS1_NOT_MIPS2 {
+ list<Predicate> InsnPredicates = [NotMips2];
+ list<Predicate> EncodingPredicates = [HasStdEnc];
+}
class ISA_MIPS1_NOT_MIPS3 {
list<Predicate> InsnPredicates = [NotMips3];
list<Predicate> EncodingPredicates = [HasStdEnc];
diff --git a/llvm/lib/Target/Mips/MipsSEISelLowering.cpp b/llvm/lib/Target/Mips/MipsSEISelLowering.cpp
index a4faefe1f65c33..9c6065332826ed 100644
--- a/llvm/lib/Target/Mips/MipsSEISelLowering.cpp
+++ b/llvm/lib/Target/Mips/MipsSEISelLowering.cpp
@@ -1274,6 +1274,10 @@ MipsSETargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI,
return MipsTargetLowering::EmitInstrWithCustomInserter(MI, BB);
case Mips::BPOSGE32_PSEUDO:
return emitBPOSGE32(MI, BB);
+ case Mips::PseudoTRUNC_W_S_M1:
+ return emitTRUNC_W_M1(MI, BB, Mips::CVT_W_S);
+ case Mips::PseudoTRUNC_W_D32_M1:
+ return emitTRUNC_W_M1(MI, BB, Mips::CVT_W_D32);
case Mips::SNZ_B_PSEUDO:
return emitMSACBranchPseudo(MI, BB, Mips::BNZ_B);
case Mips::SNZ_H_PSEUDO:
@@ -3238,6 +3242,40 @@ SDValue MipsSETargetLowering::lowerVECTOR_SHUFFLE(SDValue Op,
return lowerVECTOR_SHUFFLE_VSHF(Op, ResTy, Indices, false, DAG);
}
+// MIPS I has no trunc.w.fmt. Do the conversion with cvt.w.fmt and the
+// rounding mode set to round toward zero, then restore the FCSR. This is the
+// same sequence the assembler uses for the trunc.w.fmt macro.
+//
+// cfc1 $old, $31
+// ori $tmp, $old, 3
+// xori $tmp, $tmp, 2
+// ctc1 $tmp, $31
+// cvt.w.fmt $dst, $src
+// ctc1 $old, $31
+MachineBasicBlock *MipsSETargetLowering::emitTRUNC_W_M1(MachineInstr &MI,
+ MachineBasicBlock *BB,
+ unsigned CvtOpc) const {
+ MachineRegisterInfo &RegInfo = BB->getParent()->getRegInfo();
+ const TargetInstrInfo *TII = Subtarget.getInstrInfo();
+ const TargetRegisterClass *RC = &Mips::GPR32RegClass;
+ DebugLoc DL = MI.getDebugLoc();
+ Register Dst = MI.getOperand(0).getReg();
+ Register Src = MI.getOperand(1).getReg();
+ Register Old = RegInfo.createVirtualRegister(RC);
+ Register Tmp0 = RegInfo.createVirtualRegister(RC);
+ Register Tmp1 = RegInfo.createVirtualRegister(RC);
+
+ BuildMI(*BB, MI, DL, TII->get(Mips::CFC1), Old).addReg(Mips::FCR31);
+ BuildMI(*BB, MI, DL, TII->get(Mips::ORi), Tmp0).addReg(Old).addImm(3);
+ BuildMI(*BB, MI, DL, TII->get(Mips::XORi), Tmp1).addReg(Tmp0).addImm(2);
+ BuildMI(*BB, MI, DL, TII->get(Mips::CTC1), Mips::FCR31).addReg(Tmp1);
+ BuildMI(*BB, MI, DL, TII->get(CvtOpc), Dst).addReg(Src);
+ BuildMI(*BB, MI, DL, TII->get(Mips::CTC1), Mips::FCR31).addReg(Old);
+
+ MI.eraseFromParent();
+ return BB;
+}
+
MachineBasicBlock *
MipsSETargetLowering::emitBPOSGE32(MachineInstr &MI,
MachineBasicBlock *BB) const {
diff --git a/llvm/lib/Target/Mips/MipsSEISelLowering.h b/llvm/lib/Target/Mips/MipsSEISelLowering.h
index 18fd0f0ac84eb8..5f88d89f2ec2a3 100644
--- a/llvm/lib/Target/Mips/MipsSEISelLowering.h
+++ b/llvm/lib/Target/Mips/MipsSEISelLowering.h
@@ -94,6 +94,8 @@ using TargetRegisterClass = MCRegisterClass;
SDValue lowerFP16_TO_FP(SDValue Op, SelectionDAG &DAG) const;
SDValue lowerFP_TO_FP16(SDValue Op, SelectionDAG &DAG) const;
+ MachineBasicBlock *emitTRUNC_W_M1(MachineInstr &MI, MachineBasicBlock *BB,
+ unsigned CvtOpc) const;
MachineBasicBlock *emitBPOSGE32(MachineInstr &MI,
MachineBasicBlock *BB) const;
MachineBasicBlock *emitMSACBranchPseudo(MachineInstr &MI,
diff --git a/llvm/test/CodeGen/Mips/mips1-fptosi.ll b/llvm/test/CodeGen/Mips/mips1-fptosi.ll
new file mode 100644
index 00000000000000..f6ec8ead1ca817
--- /dev/null
+++ b/llvm/test/CodeGen/Mips/mips1-fptosi.ll
@@ -0,0 +1,91 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=mipsel-linux-gnu -mcpu=mips1 -relocation-model=static \
+; RUN: -verify-machineinstrs < %s | FileCheck %s -check-prefix=MIPS1
+; RUN: llc -mtriple=mipsel-linux-gnu -mcpu=mips2 -relocation-model=static \
+; RUN: -verify-machineinstrs < %s | FileCheck %s -check-prefix=MIPS2
+
+; MIPS I has no trunc.w.fmt. fptosi is done with cvt.w.fmt and the rounding
+; mode set to round toward zero.
+
+define i32 @fptosi_f32(float %x) {
+; MIPS1-LABEL: fptosi_f32:
+; MIPS1: # %bb.0: # %entry
+; MIPS1-NEXT: cfc1 $1, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: ori $2, $1, 3
+; MIPS1-NEXT: xori $2, $2, 2
+; MIPS1-NEXT: ctc1 $2, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: cvt.w.s $f0, $f12
+; MIPS1-NEXT: ctc1 $1, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: mfc1 $2, $f0
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: jr $ra
+; MIPS1-NEXT: nop
+;
+; MIPS2-LABEL: fptosi_f32:
+; MIPS2: # %bb.0: # %entry
+; MIPS2-NEXT: trunc.w.s $f0, $f12
+; MIPS2-NEXT: mfc1 $2, $f0
+; MIPS2-NEXT: nop
+; MIPS2-NEXT: jr $ra
+; MIPS2-NEXT: nop
+entry:
+ %r = fptosi float %x to i32
+ ret i32 %r
+}
+
+define i32 @fptosi_f64(double %x) {
+; MIPS1-LABEL: fptosi_f64:
+; MIPS1: # %bb.0: # %entry
+; MIPS1-NEXT: cfc1 $1, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: ori $2, $1, 3
+; MIPS1-NEXT: xori $2, $2, 2
+; MIPS1-NEXT: ctc1 $2, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: cvt.w.d $f0, $f12
+; MIPS1-NEXT: ctc1 $1, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: mfc1 $2, $f0
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: jr $ra
+; MIPS1-NEXT: nop
+;
+; MIPS2-LABEL: fptosi_f64:
+; MIPS2: # %bb.0: # %entry
+; MIPS2-NEXT: trunc.w.d $f0, $f12
+; MIPS2-NEXT: mfc1 $2, $f0
+; MIPS2-NEXT: nop
+; MIPS2-NEXT: jr $ra
+; MIPS2-NEXT: nop
+entry:
+ %r = fptosi double %x to i32
+ ret i32 %r
+}
+
+define void @fptosi_f32_store(float %x, ptr %p) {
+; MIPS1-LABEL: fptosi_f32_store:
+; MIPS1: # %bb.0: # %entry
+; MIPS1-NEXT: cfc1 $1, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: ori $2, $1, 3
+; MIPS1-NEXT: xori $2, $2, 2
+; MIPS1-NEXT: ctc1 $2, $31
+; MIPS1-NEXT: nop
+; MIPS1-NEXT: cvt.w.s $f0, $f12
+; MIPS1-NEXT: ctc1 $1, $31
+; MIPS1-NEXT: jr $ra
+; MIPS1-NEXT: swc1 $f0, 0($5)
+;
+; MIPS2-LABEL: fptosi_f32_store:
+; MIPS2: # %bb.0: # %entry
+; MIPS2-NEXT: trunc.w.s $f0, $f12
+; MIPS2-NEXT: jr $ra
+; MIPS2-NEXT: swc1 $f0, 0($5)
+entry:
+ %r = fptosi float %x to i32
+ store i32 %r, ptr %p
+ ret void
+}
>From 37d519eb0a931cdffd1d7fd032cc1011d5ff6615 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?G=C3=A1bor=20Horv=C3=A1th?= <xazax.hun at gmail.com>
Date: Sat, 21 Nov 2026 10:54:27 +0000
Subject: [PATCH 09/35] [ADT] Thread the ImutAVLFactory free list through the
nodes (#244626)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Freed nodes are kept in a vector that grows with the number of freed
nodes. Link them through their now unused left pointers instead,
removing the vector's allocation and growth.
Based on #196534 by Christoph Cullmann.
Assisted-by: Claude Opus 5.5
---------
Co-authored-by: Gabor Horvath <gaborh at apple.com>
Co-authored-by: Christoph Cullmann <cullmann at kde.org>
Co-authored-by: Balázs Benics <benicsbalazs at gmail.com>
---
clang/unittests/StaticAnalyzer/StoreTest.cpp | 49 +++++++++++---------
llvm/include/llvm/ADT/ImmutableSet.h | 15 ++++--
2 files changed, 38 insertions(+), 26 deletions(-)
diff --git a/clang/unittests/StaticAnalyzer/StoreTest.cpp b/clang/unittests/StaticAnalyzer/StoreTest.cpp
index d4c743c6d5a528..5b400522d0dfff 100644
--- a/clang/unittests/StaticAnalyzer/StoreTest.cpp
+++ b/clang/unittests/StaticAnalyzer/StoreTest.cpp
@@ -63,39 +63,46 @@ class VariableBindConsumer : public StoreTestConsumer {
Loc LX1 = loc::MemRegionVal(MRManager.getVarRegion(VDX1, SF));
Loc LY1 = loc::MemRegionVal(MRManager.getVarRegion(VDY1, SF));
- Store StInit = SManager.getInitialStore(SF).getStore();
+ // Keep the StoreRefs alive, a raw Store does not keep its bindings alive.
+ StoreRef StInit = SManager.getInitialStore(SF);
SVal Zero = Builder.makeZeroVal(ASTCtxt.IntTy);
SVal One = Builder.makeIntVal(1, ASTCtxt.IntTy);
SVal NarrowZero = Builder.makeZeroVal(ASTCtxt.CharTy);
// Bind(Zero)
- Store StX0 = SManager.Bind(StInit, LX0, Zero).ResultingStore.getStore();
- EXPECT_EQ(Zero, SManager.getBinding(StX0, LX0, ASTCtxt.IntTy));
+ StoreRef StX0 = SManager.Bind(StInit.getStore(), LX0, Zero).ResultingStore;
+ EXPECT_EQ(Zero, SManager.getBinding(StX0.getStore(), LX0, ASTCtxt.IntTy));
// BindDefaultInitial(Zero)
- Store StY0 = SManager.BindDefaultInitial(StInit, LY0.getAsRegion(), Zero)
- .ResultingStore.getStore();
- EXPECT_EQ(Zero, SManager.getBinding(StY0, LY0, ASTCtxt.IntTy));
- EXPECT_EQ(Zero, *SManager.getDefaultBinding(StY0, LY0.getAsRegion()));
+ StoreRef StY0 =
+ SManager.BindDefaultInitial(StInit.getStore(), LY0.getAsRegion(), Zero)
+ .ResultingStore;
+ EXPECT_EQ(Zero, SManager.getBinding(StY0.getStore(), LY0, ASTCtxt.IntTy));
+ EXPECT_EQ(Zero,
+ *SManager.getDefaultBinding(StY0.getStore(), LY0.getAsRegion()));
// BindDefaultZero()
- Store StZ0 = SManager.BindDefaultZero(StInit, LZ0.getAsRegion())
- .ResultingStore.getStore();
+ StoreRef StZ0 =
+ SManager.BindDefaultZero(StInit.getStore(), LZ0.getAsRegion())
+ .ResultingStore;
// BindDefaultZero wipes the region with '0 S8b', not with out Zero.
// Direct load, however, does give us back the object of the type
// that we specify for loading.
- EXPECT_EQ(Zero, SManager.getBinding(StZ0, LZ0, ASTCtxt.IntTy));
- EXPECT_EQ(NarrowZero, *SManager.getDefaultBinding(StZ0, LZ0.getAsRegion()));
+ EXPECT_EQ(Zero, SManager.getBinding(StZ0.getStore(), LZ0, ASTCtxt.IntTy));
+ EXPECT_EQ(NarrowZero,
+ *SManager.getDefaultBinding(StZ0.getStore(), LZ0.getAsRegion()));
// Bind(One)
- Store StX1 = SManager.Bind(StInit, LX1, One).ResultingStore.getStore();
- EXPECT_EQ(One, SManager.getBinding(StX1, LX1, ASTCtxt.IntTy));
+ StoreRef StX1 = SManager.Bind(StInit.getStore(), LX1, One).ResultingStore;
+ EXPECT_EQ(One, SManager.getBinding(StX1.getStore(), LX1, ASTCtxt.IntTy));
// BindDefaultInitial(One)
- Store StY1 = SManager.BindDefaultInitial(StInit, LY1.getAsRegion(), One)
- .ResultingStore.getStore();
- EXPECT_EQ(One, SManager.getBinding(StY1, LY1, ASTCtxt.IntTy));
- EXPECT_EQ(One, *SManager.getDefaultBinding(StY1, LY1.getAsRegion()));
+ StoreRef StY1 =
+ SManager.BindDefaultInitial(StInit.getStore(), LY1.getAsRegion(), One)
+ .ResultingStore;
+ EXPECT_EQ(One, SManager.getBinding(StY1.getStore(), LY1, ASTCtxt.IntTy));
+ EXPECT_EQ(One,
+ *SManager.getDefaultBinding(StY1.getStore(), LY1.getAsRegion()));
}
public:
@@ -132,14 +139,14 @@ class LiteralCompoundConsumer : public StoreTestConsumer {
loc::MemRegionVal ZeroElement(
MRManager.getElementRegion(ASTCtxt.IntTy, Zero, CLRegion, ASTCtxt));
- Store StInit = SManager.getInitialStore(SF).getStore();
+ StoreRef StInit = SManager.getInitialStore(SF);
// Let's bind constant 1 to 'test[0]'
SVal One = Builder.makeIntVal(1, Int);
- Store StX =
- SManager.Bind(StInit, ZeroElement, One).ResultingStore.getStore();
+ StoreRef StX =
+ SManager.Bind(StInit.getStore(), ZeroElement, One).ResultingStore;
// And make sure that we can read this binding back as it was
- EXPECT_EQ(One, SManager.getBinding(StX, ZeroElement, Int));
+ EXPECT_EQ(One, SManager.getBinding(StX.getStore(), ZeroElement, Int));
}
public:
diff --git a/llvm/include/llvm/ADT/ImmutableSet.h b/llvm/include/llvm/ADT/ImmutableSet.h
index ad80150fc2e529..d57034b7901745 100644
--- a/llvm/include/llvm/ADT/ImmutableSet.h
+++ b/llvm/include/llvm/ADT/ImmutableSet.h
@@ -374,7 +374,10 @@ class ImutAVLTree
// We need to clear the mutability bit in case we are
// destroying the node as part of a sweep in ImutAVLFactory::recoverNodes().
IsMutable = false;
- factory->freeNodes.push_back(this);
+ // The children were released above, so reuse the left pointer to link
+ // this node into the factory's free list.
+ left = factory->freeNodes;
+ factory->freeNodes = this;
}
};
@@ -405,7 +408,8 @@ class ImutAVLFactory
uintptr_t Allocator;
std::vector<TreeTy*> createdNodes;
- std::vector<TreeTy*> freeNodes;
+ // Destroyed nodes, linked through their left pointers.
+ TreeTy *freeNodes = nullptr;
bool ownsAllocator() const {
return (Allocator & 0x1) == 0;
@@ -513,9 +517,10 @@ class ImutAVLFactory
TreeTy* createNode(TreeTy* L, value_type_ref V, TreeTy* R) {
BumpPtrAllocator& A = getAllocator();
TreeTy* T;
- if (!freeNodes.empty()) {
- T = freeNodes.back();
- freeNodes.pop_back();
+ if (freeNodes) {
+ T = freeNodes;
+ freeNodes = T->getLeft();
+ assert(!T->isMutable() && "free-list node must have been destroyed");
assert(T != L);
assert(T != R);
} else {
>From aefcda1dace76c4064254d2b864c2170e036f3ff Mon Sep 17 00:00:00 2001
From: Bill Wendling <isanbard at gmail.com>
Date: Sat, 21 Nov 2026 03:21:19 -0800
Subject: [PATCH 10/35] [MIR] Serialize the operand count of inline asm operand
flags (#235184)
An inline asm operand flag is followed by the number of operands that
it says: e.g. two registers for an i128 or an i386 `"A"` output, or the
five operands of an x86 memory reference. The symbolic flag syntax had
no way to write that count, so the printer dropped it and the parser
assumed one. Printed MIR with such a flag didn't parse back correctly:
the operands after the first were read as further flags or as stray operands,
which the machine verifier rejects. Tests had to write such a flag as an
immediate, which holds its register class's TableGen ID: adding a
register class can change the immediate, though the class stays the same.
Print the count as 'ops:N' at the end of the flag when it isn't 1, and
parse it back, e.g.:
INLINEASM &"", attdialect, mem:m ops:5, %stack.0, 1, $noreg, 0, $noreg
INLINEASM &"", attdialect, reguse:GR64 foldable ops:2, %0, %1
A count of 0 is accepted, since the printer emits whatever count the
flag holds; `peephole-copy.mir` writes `1 /* clobber */`, which is a `reguse`
flag with no operands.
The existing test updates are all multi-operand flags that previously
printed without their count (x86, SystemZ, and RISC-V memory operands,
and an i386 `eax:edx` output).
The flag of a function operand (kind Func) doesn't round-trip either,
for a different reason; #237461, stacked on this PR, fixes that.
Fixes #235560.
Assisted-by: Claude Code
---------
Co-authored-by: Matt Arsenault <Matthew.Arsenault at amd.com>
---
llvm/include/llvm/IR/InlineAsm.h | 16 ++++++++-
llvm/lib/CodeGen/MIRParser/MIParser.cpp | 35 ++++++++++++++++++-
llvm/lib/CodeGen/MIRPrinter.cpp | 6 ++++
.../MIR/Generic/inline-asm-ops-errors.mir | 31 ++++++++++++++++
.../MIR/X86/inline-asm-operand-count.mir | 22 ++++++++++++
.../MIR/X86/inline-asm-rm-exhaustion.mir | 6 ++--
llvm/test/CodeGen/RISCV/rvv/pr99782.ll | 2 +-
.../SystemZ/scavenge-clobbered-reg.mir | 2 +-
.../callbr-asm-outputs-indirect-isel-m32.ll | 2 +-
llvm/test/CodeGen/X86/peephole-copy.mir | 4 +--
10 files changed, 116 insertions(+), 10 deletions(-)
create mode 100644 llvm/test/CodeGen/MIR/Generic/inline-asm-ops-errors.mir
create mode 100644 llvm/test/CodeGen/MIR/X86/inline-asm-operand-count.mir
diff --git a/llvm/include/llvm/IR/InlineAsm.h b/llvm/include/llvm/IR/InlineAsm.h
index 63490729d46ed0..7e575656038cb9 100644
--- a/llvm/include/llvm/IR/InlineAsm.h
+++ b/llvm/include/llvm/IR/InlineAsm.h
@@ -21,6 +21,7 @@
#include "llvm/IR/Value.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/MathExtras.h"
#include <cassert>
#include <string>
#include <vector>
@@ -327,7 +328,7 @@ class InlineAsm final : public Value {
explicit Flag(uint32_t F) : Storage(F) {}
Flag(enum Kind K, unsigned NumOps) : Storage(0) {
Bitfield::set<KindField>(Storage, K);
- Bitfield::set<NumOperands>(Storage, NumOps);
+ setNumOperandRegisters(NumOps);
}
operator uint32_t() { return Storage; }
Kind getKind() const { return Bitfield::get<KindField>(Storage); }
@@ -365,6 +366,19 @@ class InlineAsm final : public Value {
return Bitfield::get<NumOperands>(Storage);
}
+ /// The largest number of operands that a flag can describe.
+ static constexpr unsigned getMaxNumOperandRegisters() {
+ return maskTrailingOnes<unsigned>(NumOperands::Bits);
+ }
+
+ /// Set the number of operands that follow the flag. (The MIR parser reads
+ /// the count last, so it can't pass it to the constructor.)
+ void setNumOperandRegisters(unsigned NumOps) {
+ assert(NumOps <= getMaxNumOperandRegisters() &&
+ "Too many operands for a flag");
+ Bitfield::set<NumOperands>(Storage, NumOps);
+ }
+
/// isUseOperandTiedToDef - Return true if the flag of the inline asm
/// operand indicates it is an use operand that's matched to a def operand.
bool isUseOperandTiedToDef(unsigned &Idx) const {
diff --git a/llvm/lib/CodeGen/MIRParser/MIParser.cpp b/llvm/lib/CodeGen/MIRParser/MIParser.cpp
index 1aaf3015c9d4c9..d117f7898b53cf 100644
--- a/llvm/lib/CodeGen/MIRParser/MIParser.cpp
+++ b/llvm/lib/CodeGen/MIRParser/MIParser.cpp
@@ -454,6 +454,7 @@ class MIParser {
std::optional<unsigned> &TiedDefIdx,
bool IsDef = false);
bool parseImmediateOperand(MachineOperand &Dest);
+ bool parseInlineAsmOperandCount(InlineAsm::Flag &F);
bool parseSymbolicInlineAsmOperand(unsigned OpIdx, MachineOperand &Dest);
bool parseIRConstant(StringRef::iterator Loc, StringRef StringValue,
const Constant *&C);
@@ -1805,6 +1806,32 @@ bool MIParser::parseImmediateOperand(MachineOperand &Dest) {
return false;
}
+/// Parse the optional "ops:N" that ends an inline asm operand flag, giving the
+/// number of operands that follow it. It's omitted when there's one operand.
+bool MIParser::parseInlineAsmOperandCount(InlineAsm::Flag &F) {
+ if (Token.isNot(MIToken::Identifier) || Token.stringValue() != "ops")
+ return false;
+
+ lex();
+ if (Token.isNot(MIToken::colon))
+ return error("expected ':' after 'ops'");
+
+ lex();
+ if (Token.isNot(MIToken::IntegerLiteral))
+ return error("expected an operand count after 'ops:'");
+
+ unsigned NumOps;
+ if (getUnsigned(NumOps))
+ return true;
+
+ if (NumOps > InlineAsm::Flag::getMaxNumOperandRegisters())
+ return error("invalid inline asm operand count");
+
+ F.setNumOperandRegisters(NumOps);
+ lex();
+ return false;
+}
+
bool MIParser::parseSymbolicInlineAsmOperand(unsigned OpIdx,
MachineOperand &Dest) {
assert(OpIdx >= InlineAsm::MIOp_ExtraInfo);
@@ -1840,7 +1867,8 @@ bool MIParser::parseSymbolicInlineAsmOperand(unsigned OpIdx,
return false;
}
- // Parse symbolic form: kind[:constraint].
+ // Parse symbolic form: kind[:constraint] [foldable] [ops:N], or, for a
+ // tied use, kind tiedto:$N [ops:N].
StringRef KindStr = Token.stringValue();
constexpr auto InvalidKind = static_cast<InlineAsm::Kind>(0);
InlineAsm::Kind K =
@@ -1874,6 +1902,8 @@ bool MIParser::parseSymbolicInlineAsmOperand(unsigned OpIdx,
lex();
F.setMatchingOp(OperandNo);
+ if (parseInlineAsmOperandCount(F))
+ return true;
Dest = MachineOperand::CreateImm(F);
return false;
@@ -1947,6 +1977,9 @@ bool MIParser::parseSymbolicInlineAsmOperand(unsigned OpIdx,
lex();
}
+ if (parseInlineAsmOperandCount(F))
+ return true;
+
Dest = MachineOperand::CreateImm(F);
return false;
}
diff --git a/llvm/lib/CodeGen/MIRPrinter.cpp b/llvm/lib/CodeGen/MIRPrinter.cpp
index 1a4d6754e88ddf..7a4958c6073649 100644
--- a/llvm/lib/CodeGen/MIRPrinter.cpp
+++ b/llvm/lib/CodeGen/MIRPrinter.cpp
@@ -1030,6 +1030,12 @@ static void printMIOperand(raw_ostream &OS, MFPrintState &State,
if (IsReg && F.getRegMayBeFolded())
OS << " foldable";
+ // The count of operands that follow the flag, e.g. the two registers
+ // of an i128 or the five operands of an x86 memory reference, is
+ // needed to parse the flag back. Most flags have one, so it's implied.
+ if (unsigned NumOps = F.getNumOperandRegisters(); NumOps != 1)
+ OS << " ops:" << NumOps;
+
break;
}
}
diff --git a/llvm/test/CodeGen/MIR/Generic/inline-asm-ops-errors.mir b/llvm/test/CodeGen/MIR/Generic/inline-asm-ops-errors.mir
new file mode 100644
index 00000000000000..8c338a456a5e7b
--- /dev/null
+++ b/llvm/test/CodeGen/MIR/Generic/inline-asm-ops-errors.mir
@@ -0,0 +1,31 @@
+# RUN: split-file %s %t
+
+;--- missing-colon.mir
+# RUN: not llc -run-pass=none -filetype=null %t/missing-colon.mir 2>&1 | FileCheck %t/missing-colon.mir --check-prefix=MISSING-COLON
+---
+name: foo
+body: |
+ bb.0:
+ ; MISSING-COLON: [[@LINE+1]]:42: expected ':' after 'ops'
+ INLINEASM &"", attdialect, mem:m ops 5
+...
+
+;--- missing-count.mir
+# RUN: not llc -run-pass=none -filetype=null %t/missing-count.mir 2>&1 | FileCheck %t/missing-count.mir --check-prefix=MISSING-COUNT
+---
+name: foo
+body: |
+ bb.0:
+ ; MISSING-COUNT: [[@LINE+1]]:42: expected an operand count after 'ops:'
+ INLINEASM &"", attdialect, mem:m ops:five
+...
+
+;--- too-large.mir
+# RUN: not llc -run-pass=none -filetype=null %t/too-large.mir 2>&1 | FileCheck %t/too-large.mir --check-prefix=TOO-LARGE
+---
+name: foo
+body: |
+ bb.0:
+ ; TOO-LARGE: [[@LINE+1]]:42: invalid inline asm operand count
+ INLINEASM &"", attdialect, mem:m ops:8192
+...
diff --git a/llvm/test/CodeGen/MIR/X86/inline-asm-operand-count.mir b/llvm/test/CodeGen/MIR/X86/inline-asm-operand-count.mir
new file mode 100644
index 00000000000000..053d241aaca115
--- /dev/null
+++ b/llvm/test/CodeGen/MIR/X86/inline-asm-operand-count.mir
@@ -0,0 +1,22 @@
+# RUN: llc -mtriple=x86_64 -run-pass=none -o - %s | FileCheck %s
+
+# Test that an inline asm operand count other than 1 round-trips as "ops:N".
+
+---
+name: operand_count
+tracksRegLiveness: true
+stack:
+ - { id: 0, size: 16, alignment: 16 }
+body: |
+ bb.0:
+ liveins: $rdi, $rsi
+
+ ; CHECK-LABEL: name: operand_count
+ ; CHECK: INLINEASM &"", sideeffect mayload attdialect, regdef:GR64 ops:2, def %2, def %3, mem:m ops:5, %stack.0, 1, $noreg, 0, $noreg, reguse:GR64 foldable ops:2, %0, %1, reguse tiedto:$0 ops:2, %2(tied-def 3), %3(tied-def 4)
+ %0:gr64 = COPY $rdi
+ %1:gr64 = COPY $rsi
+ %2:gr64 = COPY %0
+ %3:gr64 = COPY %1
+ INLINEASM &"", sideeffect mayload attdialect, regdef:GR64 ops:2, def %2, def %3, mem:m ops:5, %stack.0, 1, $noreg, 0, $noreg, reguse:GR64 foldable ops:2, %0, %1, reguse tiedto:$0 ops:2, %2(tied-def 3), %3(tied-def 4)
+ RET 0, implicit %2, implicit %3
+...
diff --git a/llvm/test/CodeGen/MIR/X86/inline-asm-rm-exhaustion.mir b/llvm/test/CodeGen/MIR/X86/inline-asm-rm-exhaustion.mir
index 0fd8a6ad9de84e..99ab53f2839df6 100644
--- a/llvm/test/CodeGen/MIR/X86/inline-asm-rm-exhaustion.mir
+++ b/llvm/test/CodeGen/MIR/X86/inline-asm-rm-exhaustion.mir
@@ -81,7 +81,7 @@ machineFunctionInfo: {}
body: |
bb.0 (%ir-block.1):
; CHECK-LABEL: name: input
- ; CHECK: INLINEASM &"# $0", mayload attdialect, mem:m, %fixed-stack.0, 1, $noreg, 0, $noreg, clobber, implicit-def dead early-clobber $ax, clobber, implicit-def dead early-clobber $cx, clobber, implicit-def dead early-clobber $dx, clobber, implicit-def dead early-clobber $si, clobber, implicit-def dead early-clobber $di, clobber, implicit-def dead early-clobber $bx, clobber, implicit-def dead early-clobber $bp :: (load (s32) from %fixed-stack.0, align 16)
+ ; CHECK: INLINEASM &"# $0", mayload attdialect, mem:m ops:5, %fixed-stack.0, 1, $noreg, 0, $noreg, clobber, implicit-def dead early-clobber $ax, clobber, implicit-def dead early-clobber $cx, clobber, implicit-def dead early-clobber $dx, clobber, implicit-def dead early-clobber $si, clobber, implicit-def dead early-clobber $di, clobber, implicit-def dead early-clobber $bx, clobber, implicit-def dead early-clobber $bp :: (load (s32) from %fixed-stack.0, align 16)
; CHECK-NEXT: RET 0
%0:gr32 = MOV32rm %fixed-stack.0, 1, $noreg, 0, $noreg :: (load (s32) from %fixed-stack.0, align 16)
INLINEASM &"# $0", 0 /* attdialect */, 1076101129 /* reguse:GR32 spillable */, %0, 12 /* clobber */, implicit-def early-clobber $ax, 12 /* clobber */, implicit-def early-clobber $cx, 12 /* clobber */, implicit-def early-clobber $dx, 12 /* clobber */, implicit-def early-clobber $si, 12 /* clobber */, implicit-def early-clobber $di, 12 /* clobber */, implicit-def early-clobber $bx, 12 /* clobber */, implicit-def early-clobber $bp
@@ -141,7 +141,7 @@ machineFunctionInfo: {}
body: |
bb.0 (%ir-block.0):
; CHECK-LABEL: name: output
- ; CHECK: INLINEASM &"# $0", maystore attdialect, mem:m, %stack.1, 1, $noreg, 0, $noreg, clobber, implicit-def dead early-clobber $ax, clobber, implicit-def dead early-clobber $cx, clobber, implicit-def dead early-clobber $dx, clobber, implicit-def dead early-clobber $si, clobber, implicit-def dead early-clobber $di, clobber, implicit-def dead early-clobber $bx, clobber, implicit-def dead early-clobber $bp :: (store (s32) into %stack.1)
+ ; CHECK: INLINEASM &"# $0", maystore attdialect, mem:m ops:5, %stack.1, 1, $noreg, 0, $noreg, clobber, implicit-def dead early-clobber $ax, clobber, implicit-def dead early-clobber $cx, clobber, implicit-def dead early-clobber $dx, clobber, implicit-def dead early-clobber $si, clobber, implicit-def dead early-clobber $di, clobber, implicit-def dead early-clobber $bx, clobber, implicit-def dead early-clobber $bp :: (store (s32) into %stack.1)
; CHECK-NEXT: [[MOV32rm:%[0-9]+]]:gr32 = MOV32rm %stack.1, 1, $noreg, 0, $noreg :: (load (s32) from %stack.1)
; CHECK-NEXT: MOV32mr %stack.0, 1, $noreg, 0, $noreg, [[MOV32rm]] :: (store (s32) into %ir.1)
; CHECK-NEXT: $eax = COPY [[MOV32rm]]
@@ -208,7 +208,7 @@ body: |
; CHECK-LABEL: name: inout
; CHECK: [[MOV32rm:%[0-9]+]]:gr32 = MOV32rm %fixed-stack.0, 1, $noreg, 0, $noreg :: (load (s32) from %fixed-stack.0, align 16)
; CHECK-NEXT: MOV32mr %stack.0, 1, $noreg, 0, $noreg, [[MOV32rm]] :: (store (s32) into %stack.0)
- ; CHECK-NEXT: INLINEASM &"# $0 $1", mayload maystore attdialect, mem:m, %stack.0, 1, $noreg, 0, $noreg, mem:m, %stack.0, 1, $noreg, 0, $noreg, clobber, implicit-def dead early-clobber $ax, clobber, implicit-def dead early-clobber $cx, clobber, implicit-def dead early-clobber $dx, clobber, implicit-def dead early-clobber $si, clobber, implicit-def dead early-clobber $di, clobber, implicit-def dead early-clobber $bx, clobber, implicit-def dead early-clobber $bp :: (load store (s32) on %stack.0)
+ ; CHECK-NEXT: INLINEASM &"# $0 $1", mayload maystore attdialect, mem:m ops:5, %stack.0, 1, $noreg, 0, $noreg, mem:m ops:5, %stack.0, 1, $noreg, 0, $noreg, clobber, implicit-def dead early-clobber $ax, clobber, implicit-def dead early-clobber $cx, clobber, implicit-def dead early-clobber $dx, clobber, implicit-def dead early-clobber $si, clobber, implicit-def dead early-clobber $di, clobber, implicit-def dead early-clobber $bx, clobber, implicit-def dead early-clobber $bp :: (load store (s32) on %stack.0)
; CHECK-NEXT: [[MOV32rm1:%[0-9]+]]:gr32 = MOV32rm %stack.0, 1, $noreg, 0, $noreg :: (load (s32) from %stack.0)
; CHECK-NEXT: MOV32mr %fixed-stack.0, 1, $noreg, 0, $noreg, [[MOV32rm1]] :: (store (s32) into %ir.2, align 16)
; CHECK-NEXT: $eax = COPY [[MOV32rm1]]
diff --git a/llvm/test/CodeGen/RISCV/rvv/pr99782.ll b/llvm/test/CodeGen/RISCV/rvv/pr99782.ll
index 4d91649b55664b..90dbb1664f41b2 100644
--- a/llvm/test/CodeGen/RISCV/rvv/pr99782.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/pr99782.ll
@@ -8,7 +8,7 @@ define void @vslidedown() {
; CHECK-NEXT: [[PseudoVLE8_V_M8_:%[0-9]+]]:vrm8 = PseudoVLE8_V_M8 $noreg, [[ADDI]], -1 /* vl=VLMAX */, 3 /* e8 */, 3 /* ta, ma */ :: (load (<vscale x 1 x s512>) from %ir.v, align 1)
; CHECK-NEXT: [[ADDI1:%[0-9]+]]:gpr = ADDI %stack.1, 0
; CHECK-NEXT: PseudoVSE8_V_M8 [[PseudoVLE8_V_M8_]], [[ADDI1]], -1 /* vl=VLMAX */, 3 /* e8 */ :: (store (<vscale x 1 x s512>) into %stack.1)
- ; CHECK-NEXT: INLINEASM &"vadd.vv $0, $0, $0", sideeffect mayload maystore attdialect, mem:m, %stack.0.v, 0, mem:m, %stack.1, 0
+ ; CHECK-NEXT: INLINEASM &"vadd.vv $0, $0, $0", sideeffect mayload maystore attdialect, mem:m ops:2, %stack.0.v, 0, mem:m ops:2, %stack.1, 0
; CHECK-NEXT: PseudoRET
entry:
%v = alloca <vscale x 64 x i8>, align 1
diff --git a/llvm/test/CodeGen/SystemZ/scavenge-clobbered-reg.mir b/llvm/test/CodeGen/SystemZ/scavenge-clobbered-reg.mir
index 7ad2ef4bd8c725..dd8bbc48c289f2 100644
--- a/llvm/test/CodeGen/SystemZ/scavenge-clobbered-reg.mir
+++ b/llvm/test/CodeGen/SystemZ/scavenge-clobbered-reg.mir
@@ -27,7 +27,7 @@ body: |
; CHECK-NEXT: renamable $r0d = COPY $r3d
; CHECK-NEXT: STG killed $r2d, $r15d, 168, $noreg :: (store (s64) into %stack.1)
; CHECK-NEXT: $r2d = LAY $r15d, 4096, $noreg
- ; CHECK-NEXT: INLINEASM &" lg $0, $1\0A\09", sideeffect mayload maystore attdialect, regdef-ec:GR64Bit, def early-clobber renamable $r1d, mem:m, killed $r2d, 1080, $noreg, clobber, implicit-def dead early-clobber $r3d, clobber, implicit-def dead early-clobber $r4d, clobber, implicit-def dead early-clobber $r5d, clobber, implicit-def dead early-clobber $r6d, clobber, implicit-def dead early-clobber $r7d, clobber, implicit-def dead early-clobber $r8d, clobber, implicit-def dead early-clobber $r9d, clobber, implicit-def dead early-clobber $r10d, clobber, implicit-def dead early-clobber $r11d, clobber, implicit-def dead early-clobber $r12d, clobber, implicit-def dead early-clobber $r13d, clobber, implicit-def dead early-clobber $r14d, clobber, implicit-def dead early-clobber $cc
+ ; CHECK-NEXT: INLINEASM &" lg $0, $1\0A\09", sideeffect mayload maystore attdialect, regdef-ec:GR64Bit, def early-clobber renamable $r1d, mem:m ops:3, killed $r2d, 1080, $noreg, clobber, implicit-def dead early-clobber $r3d, clobber, implicit-def dead early-clobber $r4d, clobber, implicit-def dead early-clobber $r5d, clobber, implicit-def dead early-clobber $r6d, clobber, implicit-def dead early-clobber $r7d, clobber, implicit-def dead early-clobber $r8d, clobber, implicit-def dead early-clobber $r9d, clobber, implicit-def dead early-clobber $r10d, clobber, implicit-def dead early-clobber $r11d, clobber, implicit-def dead early-clobber $r12d, clobber, implicit-def dead early-clobber $r13d, clobber, implicit-def dead early-clobber $r14d, clobber, implicit-def dead early-clobber $cc
; CHECK-NEXT: $r2d = LG $r15d, 168, $noreg :: (load (s64) from %stack.1)
; CHECK-NEXT: renamable $r2d = AGR killed renamable $r2d, killed renamable $r0d, implicit-def dead $cc
; CHECK-NEXT: renamable $r2d = AGR killed renamable $r2d, killed renamable $r1d, implicit-def dead $cc
diff --git a/llvm/test/CodeGen/X86/callbr-asm-outputs-indirect-isel-m32.ll b/llvm/test/CodeGen/X86/callbr-asm-outputs-indirect-isel-m32.ll
index 24737ba76a6c01..1f6df4c3455837 100644
--- a/llvm/test/CodeGen/X86/callbr-asm-outputs-indirect-isel-m32.ll
+++ b/llvm/test/CodeGen/X86/callbr-asm-outputs-indirect-isel-m32.ll
@@ -87,7 +87,7 @@ define i64 @multireg() {
; CHECK: bb.0.entry:
; CHECK-NEXT: successors: %bb.1(0x80000000), %bb.2(0x00000000)
; CHECK-NEXT: {{ $}}
- ; CHECK-NEXT: INLINEASM_BR &"", attdialect, regdef, implicit-def $eax, implicit-def $edx, imm, %bb.2
+ ; CHECK-NEXT: INLINEASM_BR &"", attdialect, regdef ops:2, implicit-def $eax, implicit-def $edx, imm, %bb.2
; CHECK-NEXT: [[COPY:%[0-9]+]]:gr32 = COPY $eax
; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr32 = COPY $edx
; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr32 = COPY [[COPY1]]
diff --git a/llvm/test/CodeGen/X86/peephole-copy.mir b/llvm/test/CodeGen/X86/peephole-copy.mir
index 8ce50e59b23ded..c7b3b0fede9764 100644
--- a/llvm/test/CodeGen/X86/peephole-copy.mir
+++ b/llvm/test/CodeGen/X86/peephole-copy.mir
@@ -17,9 +17,9 @@ body: |
bb.0:
; CHECK-LABEL: name: c
; CHECK: [[MOV32ri:%[0-9]+]]:gr32_abcd = MOV32ri 512
- ; CHECK-NEXT: INLINEASM &"", sideeffect attdialect, reguse:GR32_ABCD, [[MOV32ri]], reguse, implicit-def early-clobber $df
+ ; CHECK-NEXT: INLINEASM &"", sideeffect attdialect, reguse:GR32_ABCD, [[MOV32ri]], reguse ops:0, implicit-def early-clobber $df
; CHECK-NEXT: [[MOV32ri1:%[0-9]+]]:gr32_abcd = MOV32ri 512
- ; CHECK-NEXT: INLINEASM &"", sideeffect attdialect, reguse:GR32_ABCD, [[MOV32ri1]], reguse, implicit-def early-clobber $df
+ ; CHECK-NEXT: INLINEASM &"", sideeffect attdialect, reguse:GR32_ABCD, [[MOV32ri1]], reguse ops:0, implicit-def early-clobber $df
; CHECK-NEXT: RET 0
%2:gr32 = MOV32ri 512
%0 = COPY %2
>From 6fc3a9a08759c629c743f59bc043bbcd015e1da3 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Sat, 21 Nov 2026 11:30:59 +0000
Subject: [PATCH 11/35] [VPlan] Only narrow store groups storing a single load
group (#244730)
When narrowing a store group whose members are all load interleave group
values, only the defining group of each member was checked to hold it at
the matching index. Members could come from different load groups, but
the store group was narrowed to a copy of the first member's group only.
Require all members to come from the load group of member 0.
---
.../Transforms/Vectorize/VPlanTransforms.cpp | 16 ++--
...sform-narrow-interleave-to-widen-memory.ll | 93 +++++++++++++++++++
2 files changed, 100 insertions(+), 9 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 77bec5189b11d3..6e5786e4f50bef 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -4713,15 +4713,13 @@ VPlanTransforms::narrowInterleaveGroups(VPlan &Plan,
continue;
}
- // For now, we only support full interleave groups storing load interleave
- // groups.
- if (all_of(enumerate(InterleaveR->getStoredValues()), [](auto Op) {
- VPRecipeBase *DefR = Op.value()->getDefiningRecipe();
- if (!DefR)
- return false;
- auto *IR = dyn_cast<VPInterleaveRecipe>(DefR);
- return IR && IR->getInterleaveGroup()->isFull() &&
- IR->getVPValue(Op.index()) == Op.value();
+ // For now, we only support full interleave groups storing a single full
+ // load interleave group in order.
+ auto *LoadGroup =
+ dyn_cast_or_null<VPInterleaveRecipe>(Member0->getDefiningRecipe());
+ if (LoadGroup && LoadGroup->getInterleaveGroup()->isFull() &&
+ all_of(enumerate(InterleaveR->getStoredValues()), [LoadGroup](auto Op) {
+ return LoadGroup->getVPValue(Op.index()) == Op.value();
})) {
StoreGroups.push_back(InterleaveR);
continue;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll
index 92cd0be1e7f9c3..215b0d198bec60 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll
@@ -955,3 +955,96 @@ loop:
exit:
ret void
}
+
+; The stored members come from two different load groups (field 0 of %src.0 and
+; field 1 of %src.1), so the store group cannot be narrowed to a copy of a
+; single wide load.
+define void @store_group_members_from_different_load_groups(ptr noalias %src.0, ptr noalias %src.1, ptr noalias %dst) {
+; VF2-LABEL: define void @store_group_members_from_different_load_groups(
+; VF2-SAME: ptr noalias [[SRC_0:%.*]], ptr noalias [[SRC_1:%.*]], ptr noalias [[DST:%.*]]) {
+; VF2-NEXT: [[ENTRY:.*:]]
+; VF2-NEXT: br label %[[VECTOR_PH:.*]]
+; VF2: [[VECTOR_PH]]:
+; VF2-NEXT: br label %[[VECTOR_BODY:.*]]
+; VF2: [[VECTOR_BODY]]:
+; VF2-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; VF2-NEXT: [[TMP0:%.*]] = shl nsw i64 [[INDEX]], 1
+; VF2-NEXT: [[TMP5:%.*]] = or disjoint i64 [[TMP0]], 1
+; VF2-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[SRC_0]], i64 [[TMP0]]
+; VF2-NEXT: [[WIDE_VEC:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
+; VF2-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <4 x i64> [[WIDE_VEC]], <4 x i64> poison, <2 x i32> <i32 0, i32 2>
+; VF2-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <4 x i64> [[WIDE_VEC]], <4 x i64> poison, <2 x i32> <i32 1, i32 3>
+; VF2-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[SRC_1]], i64 [[TMP5]]
+; VF2-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i64 -8
+; VF2-NEXT: [[WIDE_VEC2:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; VF2-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <4 x i64> [[WIDE_VEC2]], <4 x i64> poison, <2 x i32> <i32 0, i32 2>
+; VF2-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <4 x i64> [[WIDE_VEC2]], <4 x i64> poison, <2 x i32> <i32 1, i32 3>
+; VF2-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP0]]
+; VF2-NEXT: [[TMP6:%.*]] = shufflevector <2 x i64> [[STRIDED_VEC]], <2 x i64> [[STRIDED_VEC4]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+; VF2-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> poison, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; VF2-NEXT: store <4 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8
+; VF2-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; VF2-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; VF2-NEXT: br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; VF2: [[MIDDLE_BLOCK]]:
+; VF2-NEXT: br label %[[EXIT:.*]]
+; VF2: [[EXIT]]:
+; VF2-NEXT: ret void
+;
+; VF4-LABEL: define void @store_group_members_from_different_load_groups(
+; VF4-SAME: ptr noalias [[SRC_0:%.*]], ptr noalias [[SRC_1:%.*]], ptr noalias [[DST:%.*]]) {
+; VF4-NEXT: [[ENTRY:.*:]]
+; VF4-NEXT: br label %[[VECTOR_PH:.*]]
+; VF4: [[VECTOR_PH]]:
+; VF4-NEXT: br label %[[VECTOR_BODY:.*]]
+; VF4: [[VECTOR_BODY]]:
+; VF4-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; VF4-NEXT: [[TMP0:%.*]] = shl nsw i64 [[INDEX]], 1
+; VF4-NEXT: [[TMP1:%.*]] = or disjoint i64 [[TMP0]], 1
+; VF4-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[SRC_0]], i64 [[TMP0]]
+; VF4-NEXT: [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; VF4-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; VF4-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[SRC_1]], i64 [[TMP1]]
+; VF4-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i64 -8
+; VF4-NEXT: [[WIDE_VEC2:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8
+; VF4-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <8 x i64> [[WIDE_VEC2]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; VF4-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <8 x i64> [[WIDE_VEC2]], <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; VF4-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP0]]
+; VF4-NEXT: [[TMP6:%.*]] = shufflevector <4 x i64> [[STRIDED_VEC]], <4 x i64> [[STRIDED_VEC4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; VF4-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i64> [[TMP6]], <8 x i64> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; VF4-NEXT: store <8 x i64> [[INTERLEAVED_VEC]], ptr [[TMP5]], align 8
+; VF4-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; VF4-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; VF4-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; VF4: [[MIDDLE_BLOCK]]:
+; VF4-NEXT: br label %[[EXIT:.*]]
+; VF4: [[EXIT]]:
+; VF4-NEXT: ret void
+;
+entry:
+ br label %loop
+
+loop:
+ %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+ %mul.2 = shl nsw i64 %iv, 1
+ %add.1 = or disjoint i64 %mul.2, 1
+ %gep.src.0.0 = getelementptr inbounds i64, ptr %src.0, i64 %mul.2
+ %l.0.0 = load i64, ptr %gep.src.0.0, align 8
+ %gep.src.0.1 = getelementptr inbounds i64, ptr %src.0, i64 %add.1
+ %l.0.1 = load i64, ptr %gep.src.0.1, align 8
+ %gep.src.1.0 = getelementptr inbounds i64, ptr %src.1, i64 %mul.2
+ %l.1.0 = load i64, ptr %gep.src.1.0, align 8
+ %gep.src.1.1 = getelementptr inbounds i64, ptr %src.1, i64 %add.1
+ %l.1.1 = load i64, ptr %gep.src.1.1, align 8
+ %gep.dst.0 = getelementptr inbounds i64, ptr %dst, i64 %mul.2
+ store i64 %l.0.0, ptr %gep.dst.0, align 8
+ %gep.dst.1 = getelementptr inbounds i64, ptr %dst, i64 %add.1
+ store i64 %l.1.1, ptr %gep.dst.1, align 8
+ %iv.next = add nuw nsw i64 %iv, 1
+ %ec = icmp eq i64 %iv.next, 100
+ br i1 %ec, label %exit, label %loop
+
+exit:
+ ret void
+}
>From 5689d1bcbc57030e576c74b1ee5dd0bd936247d6 Mon Sep 17 00:00:00 2001
From: David Green <david.green at arm.com>
Date: Sat, 21 Nov 2026 11:33:39 +0000
Subject: [PATCH 12/35] [ARM] Promote bf16 fp rounding operations. (#244867)
Similar to other operations we promote bf16 fp round operations to fp32
---
llvm/lib/Target/ARM/ARMISelLowering.cpp | 4 +-
llvm/test/CodeGen/ARM/bf16-instructions.ll | 191 +++++++++++++++++----
2 files changed, 164 insertions(+), 31 deletions(-)
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index c178e740de3d36..f3d6881dfc2b08 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -584,7 +584,9 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM_,
setOperationAction(ISD::BF16_TO_FP, MVT::f32, Expand);
setOperationAction(ISD::BF16_TO_FP, MVT::f64, Expand);
- for (auto Op : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV})
+ for (auto Op : {ISD::FADD, ISD::FSUB, ISD::FMUL, ISD::FDIV, ISD::FCEIL,
+ ISD::FFLOOR, ISD::FNEARBYINT, ISD::FRINT, ISD::FROUND,
+ ISD::FROUNDEVEN, ISD::FTRUNC})
setOperationAction(Op, MVT::bf16, Promote);
} else {
setOperationAction(ISD::BF16_TO_FP, MVT::f32, Expand);
diff --git a/llvm/test/CodeGen/ARM/bf16-instructions.ll b/llvm/test/CodeGen/ARM/bf16-instructions.ll
index 43892c40118269..cadc16b4740916 100644
--- a/llvm/test/CodeGen/ARM/bf16-instructions.ll
+++ b/llvm/test/CodeGen/ARM/bf16-instructions.ll
@@ -1148,8 +1148,6 @@ define bfloat @test_copysign_f64(bfloat %a, double %b) #0 {
ret bfloat %r
}
-; away the (fpext (fp_round <result>)) here.
-
define float @test_copysign_extended(bfloat %a, bfloat %b) #0 {
; CHECK-CVT-LABEL: test_copysign_extended:
; CHECK-CVT: @ %bb.0:
@@ -1207,40 +1205,173 @@ define float @test_copysign_extended(bfloat %a, bfloat %b) #0 {
ret float %xr
}
-;define bfloat @test_floor(bfloat %a) #0 {
-; %r = call bfloat @llvm.floor.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_floor(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_floor:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrintm.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_floor:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrintm.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.floor.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_ceil(bfloat %a) #0 {
-; %r = call bfloat @llvm.ceil.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_ceil(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_ceil:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrintp.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_ceil:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrintp.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.ceil.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_trunc(bfloat %a) #0 {
-; %r = call bfloat @llvm.trunc.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_trunc(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_trunc:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrintz.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_trunc:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrintz.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.trunc.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_rint(bfloat %a) #0 {
-; %r = call bfloat @llvm.rint.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_rint(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_rint:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrintx.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_rint:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrintx.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.rint.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_nearbyint(bfloat %a) #0 {
-; %r = call bfloat @llvm.nearbyint.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_nearbyint(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_nearbyint:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrintr.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_nearbyint:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrintr.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.nearbyint.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_round(bfloat %a) #0 {
-; %r = call bfloat @llvm.round.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_round(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_round:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrinta.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_round:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrinta.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.round.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_roundeven(bfloat %a) #0 {
-; %r = call bfloat @llvm.roundeven.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_roundeven(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_roundeven:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: vrintn.f32 s0, s0
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_roundeven:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: vrintn.f32 s0, s0
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: bx lr
+ %r = call bfloat @llvm.roundeven.bf16(bfloat %a)
+ ret bfloat %r
+}
define bfloat @test_fmuladd(bfloat %a, bfloat %b, bfloat %c) #0 {
; CHECK-CVT-LABEL: test_fmuladd:
>From a5afaf1bc19da8e1b112365afc7dbe8c57e2cd01 Mon Sep 17 00:00:00 2001
From: actink <actink at 163.com>
Date: Sat, 21 Nov 2026 19:46:19 +0800
Subject: [PATCH 13/35] [PowerPC] Model the CR defs of MTCRF used to restore
callee-saved CRs (#242763)
The epilogue restores the non-volatile CR fields (CR2-CR4) with a single
`mtcrf`. MTCRF takes only an FXM mask immediate and a GPR, and declares
no CR operands, so the post-RA scheduler sees no dependency between the
restore and instructions that read or write the restored CR bits. It can
therefore hoist the restore above those reads, making the function use
the caller's condition bits instead of its own.
Add the restored CR fields as implicit defs when building the MTCRF in
`emitEpilogue()` and `restoreCRs()`, matching the way the prologue
already adds implicit uses to MFCR. The scheduler now models the WAR/WAW
dependencies and keeps the restore after the last read.
This is a regression from #202339 (3fc6711ccaf6), which replaced the
per-field MTOCRF sequence -- each MTOCRF defines the CR field it
restores -- with a single MTCRF, which declares no CR defs.
Fixes #242250.
---
llvm/lib/Target/PowerPC/PPCFrameLowering.cpp | 27 ++++++++++++++-----
llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll | 7 ++---
2 files changed, 23 insertions(+), 11 deletions(-)
diff --git a/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp b/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp
index 6d8300fc8faa26..d59c02806997b2 100644
--- a/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp
@@ -1781,9 +1781,16 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF,
for (unsigned CRField : MustSaveCRs) {
CRMask |= 0x80 >> (CRField - PPC::CR0);
}
- BuildMI(MBB, InsertPt, dl, TII.get(isPPC64 ? PPC::MTCRF8 : PPC::MTCRF))
- .addImm(CRMask)
- .addReg(SrcReg, getKillRegState(true));
+ MachineInstrBuilder MIB =
+ BuildMI(MBB, InsertPt, dl,
+ TII.get(isPPC64 ? PPC::MTCRF8 : PPC::MTCRF))
+ .addImm(CRMask)
+ .addReg(SrcReg, getKillRegState(true));
+ // MTCRF does not statically model which CR fields it writes. Add the
+ // restored fields as implicit defs so that the post-RA scheduler treats
+ // them as writes (and doesn't move MTCRF above reads of those fields).
+ for (unsigned CRField : MustSaveCRs)
+ MIB.addDef(CRField, RegState::Implicit);
}
};
@@ -2581,9 +2588,17 @@ static void restoreCRs(bool is31, bool CR2Spilled, bool CR3Spilled,
CRMask |= 0x10;
if (CR4Spilled)
CRMask |= 0x08;
- MBB.insert(MI, BuildMI(*MF, DL, TII.get(PPC::MTCRF))
- .addImm(CRMask)
- .addReg(MoveReg, getKillRegState(true)));
+ MachineInstrBuilder MIB = BuildMI(*MF, DL, TII.get(PPC::MTCRF))
+ .addImm(CRMask)
+ .addReg(MoveReg, getKillRegState(true));
+ // Add the restored CR fields as implicit defs (see BuildMoveToCR).
+ if (CR2Spilled)
+ MIB.addDef(PPC::CR2, RegState::Implicit);
+ if (CR3Spilled)
+ MIB.addDef(PPC::CR3, RegState::Implicit);
+ if (CR4Spilled)
+ MIB.addDef(PPC::CR4, RegState::Implicit);
+ MBB.insert(MI, MIB);
}
}
diff --git a/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll b/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll
index 8a0dd5d8e4e1da..8d2c859eed76fa 100644
--- a/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll
+++ b/llvm/test/CodeGen/PowerPC/mtcrf-cr-restore.ll
@@ -3,10 +3,7 @@
;
; The epilogue restores the non-volatile CR fields with mtcrf. The post-RA
; scheduler must not move that restore above reads of the CR bits it writes.
-;
-; FIXME: MTCRF does not model the CR fields it writes, so the restore is
-; currently hoisted above some CR2/CR3 reads. Update this test once MTCRF
-; models its defs. See https://github.com/llvm/llvm-project/issues/242250.
+; See https://github.com/llvm/llvm-project/issues/242250.
define signext i32 @restore_cr_after_uses(i128 %value, i1 %cond) {
; CHECK-LABEL: restore_cr_after_uses:
@@ -37,8 +34,8 @@ define signext i32 @restore_cr_after_uses(i128 %value, i1 %cond) {
; CHECK-NEXT: li 3, 1
; CHECK-NEXT: crandc 20, 8, 30
; CHECK-NEXT: lwz 12, 8(1)
-; CHECK-NEXT: mtcrf 48, 12
; CHECK-NEXT: crand 22, 30, 12
+; CHECK-NEXT: mtcrf 48, 12
; CHECK-NEXT: crnor 20, 22, 20
; CHECK-NEXT: isel 4, 0, 3, 20
; CHECK-NEXT: crandc 20, 24, 6
>From 4d3d622e99e50163981323a3d114a92006956a37 Mon Sep 17 00:00:00 2001
From: Lewis Crawford <lcrawford at nvidia.com>
Date: Sat, 21 Nov 2026 11:50:03 +0000
Subject: [PATCH 14/35] [SelectionDAG] Honor denormal modes when folding
(#236172)
Introduce shared denormal-aware FPFold helper functions in the Support
library.
Then use the new tryFoldFP API within SelectionDAG to correctly fold
arithmetic, FMA, comparisons, rounding, and min/max using the enclosing
MachineFunction's denormal mode.
Decline folds when dynamic flushing can change the result or IEEE
rounding can promote a subnormal FTZ result to minimum-normal. Update
the NVPTX tests added in #236060 to verify the correct behaviour.
---
llvm/include/llvm/CodeGen/SelectionDAG.h | 4 +
llvm/include/llvm/Support/FPFold.h | 82 ++++++++
llvm/lib/Analysis/TargetTransformInfo.cpp | 4 +
.../lib/CodeGen/SelectionDAG/SelectionDAG.cpp | 73 ++++---
llvm/lib/Support/CMakeLists.txt | 1 +
llvm/lib/Support/FPFold.cpp | 195 ++++++++++++++++++
.../CodeGen/NVPTX/denormal-constant-fold.ll | 135 +++++++-----
.../gn/secondary/llvm/lib/Support/BUILD.gn | 1 +
8 files changed, 411 insertions(+), 84 deletions(-)
create mode 100644 llvm/include/llvm/Support/FPFold.h
create mode 100644 llvm/lib/Support/FPFold.cpp
diff --git a/llvm/include/llvm/CodeGen/SelectionDAG.h b/llvm/include/llvm/CodeGen/SelectionDAG.h
index 5bd0454067c1ec..9956184358f909 100644
--- a/llvm/include/llvm/CodeGen/SelectionDAG.h
+++ b/llvm/include/llvm/CodeGen/SelectionDAG.h
@@ -40,6 +40,7 @@
#include "llvm/Support/CodeGen.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/FPFold.h"
#include "llvm/Support/KnownFPClass.h"
#include "llvm/Support/RecyclingAllocator.h"
#include "llvm/Support/UndefPoison.h"
@@ -756,6 +757,9 @@ class SelectionDAG {
bool isTarget = false);
LLVM_ABI SDValue getConstantFP(const APFloat &Val, const SDLoc &DL, EVT VT,
bool isTarget = false);
+ /// Return the folded constant, or an empty SDValue if folding failed.
+ LLVM_ABI SDValue getConstFP(std::optional<FPFoldResult> Result,
+ const SDLoc &DL, EVT VT, bool isTarget = false);
LLVM_ABI SDValue getConstantFP(const ConstantFP &V, const SDLoc &DL, EVT VT,
bool isTarget = false);
SDValue getTargetConstantFP(double Val, const SDLoc &DL, EVT VT) {
diff --git a/llvm/include/llvm/Support/FPFold.h b/llvm/include/llvm/Support/FPFold.h
new file mode 100644
index 00000000000000..7f517feb92393c
--- /dev/null
+++ b/llvm/include/llvm/Support/FPFold.h
@@ -0,0 +1,82 @@
+//===- FPFold.h - Floating-point constant-folding helpers -------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+/// \file
+/// This file provides fallible floating-point folding operations with explicit
+/// rounding and denormal modes.
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_SUPPORT_FPFOLD_H
+#define LLVM_SUPPORT_FPFOLD_H
+
+#include "llvm/ADT/APFloat.h"
+#include "llvm/ADT/ArrayRef.h"
+#include "llvm/ADT/FloatingPointMode.h"
+#include "llvm/Support/Compiler.h"
+#include <optional>
+
+namespace llvm {
+
+/// The floating-point operations supported by the folding helpers.
+enum class FPOp {
+ Add,
+ Sub,
+ Mul,
+ Div,
+ FRem,
+ MinNum,
+ MaxNum,
+ Minimum,
+ Maximum,
+ MinimumNum,
+ MaximumNum,
+ FMA,
+ Ceil,
+ Floor,
+ Trunc,
+ Round,
+ RoundEven
+};
+
+/// A unique floating-point result and the IEEE exceptions raised computing it.
+struct FPFoldResult {
+ APFloat Value;
+ APFloat::opStatus Status;
+};
+
+/// Return the value of a successful fold, or std::nullopt.
+LLVM_ABI std::optional<APFloat> getFPValue(std::optional<FPFoldResult> Res);
+
+/// Try to evaluate a floating-point operation with the specified inputs.
+/// Ceil, Floor, Trunc, Round, and RoundEven select their own rounding mode;
+/// Other rounding-dependent operations use RN.
+/// Min/max and FRem do not depend on the rounding mode.
+///
+/// Returns std::nullopt when the modes do not specify one result. In
+/// particular, this occurs when a dynamic denormal mode could change an input
+/// or output value.
+LLVM_ABI std::optional<FPFoldResult>
+tryFoldFP(FPOp Opcode, ArrayRef<APFloat> Args, DenormalMode Denorms);
+
+/// Try to evaluate a floating-point operation with an explicit rounding mode.
+/// Dynamic is evaluated using RN, but folding is declined if evaluation is
+/// inexact. Fixed-rounding operations still select their own rounding mode.
+LLVM_ABI std::optional<FPFoldResult> tryFoldFPWithRM(FPOp Opcode,
+ ArrayRef<APFloat> Args,
+ RoundingMode RM,
+ DenormalMode Denorms);
+
+/// Try to compare two floating-point values.
+///
+/// A comparison has no rounding or output-denormal mode, but is not foldable
+/// when dynamic input-denormal handling could alter an operand.
+LLVM_ABI std::optional<APFloat::cmpResult>
+tryFoldFCmp(const APFloat &LHS, const APFloat &RHS, DenormalMode Denorms);
+
+} // namespace llvm
+
+#endif // LLVM_SUPPORT_FPFOLD_H
diff --git a/llvm/lib/Analysis/TargetTransformInfo.cpp b/llvm/lib/Analysis/TargetTransformInfo.cpp
index 0a225409bf2536..2ac8d9b62dba60 100644
--- a/llvm/lib/Analysis/TargetTransformInfo.cpp
+++ b/llvm/lib/Analysis/TargetTransformInfo.cpp
@@ -808,6 +808,10 @@ InstructionCost TargetTransformInfo::getFPOpCost(Type *Ty) const {
return Cost;
}
+// Prefer constant folding to use IEEE by default, and ignore the
+// denormal_fpenv function attribute. Targets may override this to return DM
+// unmodified to match denormal_fpenv where possible when folding, or may
+// provide custom rules for specific input types / operations.
InstructionCost TargetTransformInfo::getIntImmCodeSizeCost(unsigned Opcode,
unsigned Idx,
const APInt &Imm,
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 6c335697ddb39e..4c8b0daa844682 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -66,6 +66,7 @@
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/FPFold.h"
#include "llvm/Support/KnownBits.h"
#include "llvm/Support/KnownFPClass.h"
#include "llvm/Support/MathExtras.h"
@@ -1845,6 +1846,11 @@ SDValue SelectionDAG::getConstantFP(const APFloat &V, const SDLoc &DL, EVT VT,
return getConstantFP(*ConstantFP::get(*getContext(), V), DL, VT, isTarget);
}
+SDValue SelectionDAG::getConstFP(std::optional<FPFoldResult> Result,
+ const SDLoc &DL, EVT VT, bool isTarget) {
+ return Result ? getConstantFP(Result->Value, DL, VT, isTarget) : SDValue();
+}
+
SDValue SelectionDAG::getConstantFP(const ConstantFP &V, const SDLoc &DL,
EVT VT, bool isTarget) {
assert(VT.isFloatingPoint() && "Cannot create integer FP constant!");
@@ -2758,7 +2764,11 @@ SDValue SelectionDAG::FoldSetCC(EVT VT, SDValue N1, SDValue N2,
auto *N2CFP = dyn_cast<ConstantFPSDNode>(N2);
if (N1CFP && N2CFP) {
- APFloat::cmpResult R = N1CFP->getValueAPF().compare(N2CFP->getValueAPF());
+ auto MaybeR = tryFoldFCmp(N1CFP->getValueAPF(), N2CFP->getValueAPF(),
+ getDenormalMode(OpVT));
+ if (!MaybeR)
+ return SDValue();
+ APFloat::cmpResult R = *MaybeR;
switch (Cond) {
default: break;
case ISD::SETEQ: if (R==APFloat::cmpUnordered)
@@ -7927,6 +7937,7 @@ SDValue SelectionDAG::FoldConstantArithmetic(unsigned Opcode, const SDLoc &DL,
// Constant fold unary operations with a floating point constant operand.
if (auto *C = dyn_cast<ConstantFPSDNode>(N1)) {
APFloat V = C->getValueAPF(); // make copy
+ DenormalMode DenormMode = getDenormalMode(N1.getValueType());
switch (Opcode) {
case ISD::FNEG:
V.changeSign();
@@ -7935,21 +7946,21 @@ SDValue SelectionDAG::FoldConstantArithmetic(unsigned Opcode, const SDLoc &DL,
V.clearSign();
return getConstantFP(V, DL, VT);
case ISD::FCEIL: {
- APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardPositive);
- if (fs == APFloat::opOK || fs == APFloat::opInexact)
- return getConstantFP(V, DL, VT);
+ if (auto Res = tryFoldFP(FPOp::Ceil, {V}, DenormMode))
+ if (Res->Status == APFloat::opOK || Res->Status == APFloat::opInexact)
+ return getConstantFP(Res->Value, DL, VT);
return SDValue();
}
case ISD::FTRUNC: {
- APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardZero);
- if (fs == APFloat::opOK || fs == APFloat::opInexact)
- return getConstantFP(V, DL, VT);
+ if (auto Res = tryFoldFP(FPOp::Trunc, {V}, DenormMode))
+ if (Res->Status == APFloat::opOK || Res->Status == APFloat::opInexact)
+ return getConstantFP(Res->Value, DL, VT);
return SDValue();
}
case ISD::FFLOOR: {
- APFloat::opStatus fs = V.roundToIntegral(APFloat::rmTowardNegative);
- if (fs == APFloat::opOK || fs == APFloat::opInexact)
- return getConstantFP(V, DL, VT);
+ if (auto Res = tryFoldFP(FPOp::Floor, {V}, DenormMode))
+ if (Res->Status == APFloat::opOK || Res->Status == APFloat::opInexact)
+ return getConstantFP(Res->Value, DL, VT);
return SDValue();
}
case ISD::FP_EXTEND: {
@@ -8122,15 +8133,17 @@ SDValue SelectionDAG::FoldConstantArithmetic(unsigned Opcode, const SDLoc &DL,
ConstantFPSDNode *C2 = dyn_cast<ConstantFPSDNode>(Ops[1]);
ConstantFPSDNode *C3 = dyn_cast<ConstantFPSDNode>(Ops[2]);
if (C1 && C2 && C3) {
+ DenormalMode DM = getDenormalMode(VT);
APFloat V1 = C1->getValueAPF();
const APFloat &V2 = C2->getValueAPF();
const APFloat &V3 = C3->getValueAPF();
if (Opcode == ISD::FMAD || Opcode == ISD::FMULADD) {
- V1.multiply(V2, APFloat::rmNearestTiesToEven);
- V1.add(V3, APFloat::rmNearestTiesToEven);
- } else
- V1.fusedMultiplyAdd(V2, V3, APFloat::rmNearestTiesToEven);
- return getConstantFP(V1, DL, VT);
+ if (auto Product = tryFoldFP(FPOp::Mul, {V1, V2}, DM))
+ return getConstFP(tryFoldFP(FPOp::Add, {Product->Value, V3}, DM), DL,
+ VT);
+ return SDValue();
+ }
+ return getConstFP(tryFoldFP(FPOp::FMA, {V1, V2, V3}, DM), DL, VT);
}
}
@@ -8481,39 +8494,35 @@ SDValue SelectionDAG::foldConstantFPMath(unsigned Opcode, const SDLoc &DL,
ConstantFPSDNode *N1CFP = isConstOrConstSplatFP(N1, /*AllowUndefs*/ false);
ConstantFPSDNode *N2CFP = isConstOrConstSplatFP(N2, /*AllowUndefs*/ false);
if (N1CFP && N2CFP) {
- APFloat C1 = N1CFP->getValueAPF(); // make copy
+ APFloat C1 = N1CFP->getValueAPF();
const APFloat &C2 = N2CFP->getValueAPF();
+ DenormalMode DM = getDenormalMode(VT);
switch (Opcode) {
case ISD::FADD:
- C1.add(C2, APFloat::rmNearestTiesToEven);
- return getConstantFP(C1, DL, VT);
+ return getConstFP(tryFoldFP(FPOp::Add, {C1, C2}, DM), DL, VT);
case ISD::FSUB:
- C1.subtract(C2, APFloat::rmNearestTiesToEven);
- return getConstantFP(C1, DL, VT);
+ return getConstFP(tryFoldFP(FPOp::Sub, {C1, C2}, DM), DL, VT);
case ISD::FMUL:
- C1.multiply(C2, APFloat::rmNearestTiesToEven);
- return getConstantFP(C1, DL, VT);
+ return getConstFP(tryFoldFP(FPOp::Mul, {C1, C2}, DM), DL, VT);
case ISD::FDIV:
- C1.divide(C2, APFloat::rmNearestTiesToEven);
- return getConstantFP(C1, DL, VT);
+ return getConstFP(tryFoldFP(FPOp::Div, {C1, C2}, DM), DL, VT);
case ISD::FREM:
- C1.mod(C2);
- return getConstantFP(C1, DL, VT);
+ return getConstFP(tryFoldFP(FPOp::FRem, {C1, C2}, DM), DL, VT);
case ISD::FCOPYSIGN:
C1.copySign(C2);
return getConstantFP(C1, DL, VT);
case ISD::FMINNUM:
- return getConstantFP(minnum(C1, C2), DL, VT);
+ return getConstFP(tryFoldFP(FPOp::MinNum, {C1, C2}, DM), DL, VT);
case ISD::FMAXNUM:
- return getConstantFP(maxnum(C1, C2), DL, VT);
+ return getConstFP(tryFoldFP(FPOp::MaxNum, {C1, C2}, DM), DL, VT);
case ISD::FMINIMUM:
- return getConstantFP(minimum(C1, C2), DL, VT);
+ return getConstFP(tryFoldFP(FPOp::Minimum, {C1, C2}, DM), DL, VT);
case ISD::FMAXIMUM:
- return getConstantFP(maximum(C1, C2), DL, VT);
+ return getConstFP(tryFoldFP(FPOp::Maximum, {C1, C2}, DM), DL, VT);
case ISD::FMINIMUMNUM:
- return getConstantFP(minimumnum(C1, C2), DL, VT);
+ return getConstFP(tryFoldFP(FPOp::MinimumNum, {C1, C2}, DM), DL, VT);
case ISD::FMAXIMUMNUM:
- return getConstantFP(maximumnum(C1, C2), DL, VT);
+ return getConstFP(tryFoldFP(FPOp::MaximumNum, {C1, C2}, DM), DL, VT);
default: break;
}
}
diff --git a/llvm/lib/Support/CMakeLists.txt b/llvm/lib/Support/CMakeLists.txt
index d2ec0e728bfc03..96bcd938c3b645 100644
--- a/llvm/lib/Support/CMakeLists.txt
+++ b/llvm/lib/Support/CMakeLists.txt
@@ -223,6 +223,7 @@ add_llvm_component_library(LLVMSupport
FileUtilities.cpp
FileOutputBuffer.cpp
FloatingPointMode.cpp
+ FPFold.cpp
FoldingSet.cpp
FormattedStream.cpp
FormatVariadic.cpp
diff --git a/llvm/lib/Support/FPFold.cpp b/llvm/lib/Support/FPFold.cpp
new file mode 100644
index 00000000000000..7a5dd7e3601c81
--- /dev/null
+++ b/llvm/lib/Support/FPFold.cpp
@@ -0,0 +1,195 @@
+//===- FPFold.cpp - Floating-point constant-folding helpers --------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Support/FPFold.h"
+#include "llvm/ADT/STLForwardCompat.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/Support/ErrorHandling.h"
+#include <cassert>
+#include <utility>
+
+using namespace llvm;
+
+std::optional<APFloat> llvm::getFPValue(std::optional<FPFoldResult> Res) {
+ return llvm::transformOptional(std::move(Res), &FPFoldResult::Value);
+}
+
+static APFloat flush(APFloat Value, DenormalMode::DenormalModeKind Mode) {
+ if (!Value.isDenormal())
+ return Value;
+
+ if (Mode == DenormalMode::PreserveSign)
+ return APFloat::getZero(Value.getSemantics(), Value.isNegative());
+ if (Mode == DenormalMode::PositiveZero)
+ return APFloat::getZero(Value.getSemantics());
+ return Value;
+}
+
+/// Return true if gradual-underflow rounding may have promoted a result from
+/// the subnormal range to the smallest normal value. When output denormals are
+/// flushed, folding such a result as an ordinary IEEE operation is unsafe: an
+/// FTZ operation may produce zero before gradual-underflow rounding can perform
+/// that promotion.
+static bool mayHaveBeenRoundedFromSubnormal(const APFloat &Result,
+ APFloat::opStatus Status,
+ RoundingMode RM) {
+ if (!(Status & APFloat::opInexact) || !Result.isSmallestNormalized())
+ return false;
+
+ switch (RM) {
+ default:
+ llvm_unreachable("unknown rounding mode");
+ case RoundingMode::NearestTiesToEven:
+ case RoundingMode::NearestTiesToAway:
+ case RoundingMode::Dynamic:
+ return true;
+ case RoundingMode::TowardPositive:
+ return !Result.isNegative();
+ case RoundingMode::TowardNegative:
+ return Result.isNegative();
+ case RoundingMode::TowardZero:
+ return false;
+ }
+}
+
+static APFloat::opStatus getSNaNStatus(ArrayRef<APFloat> Args) {
+ for (const APFloat &Arg : Args)
+ if (Arg.isSignaling())
+ return APFloat::opInvalidOp;
+ return APFloat::opOK;
+}
+
+/// The internal helper that dispatches to APFloat, and always produces
+/// a status and result (so RM must be valid and non-dynamic if used, and
+/// input args must already be flushed if necessary).
+/// Result must be initialized with a copy of Args[0].
+static APFloat::opStatus executeFold(FPOp Opcode, APFloat &Result,
+ ArrayRef<APFloat> Args, RoundingMode RM) {
+ assert(!Args.empty() && Result.bitwiseIsEqual(Args[0]) &&
+ "Result must start as a copy of Args[0]");
+ if (Args.size() == 2) {
+ switch (Opcode) {
+ default:
+ llvm_unreachable("unsupported 2-arg FP folding op");
+ case FPOp::Add:
+ return Result.add(Args[1], RM);
+ case FPOp::Sub:
+ return Result.subtract(Args[1], RM);
+ case FPOp::Mul:
+ return Result.multiply(Args[1], RM);
+ case FPOp::Div:
+ return Result.divide(Args[1], RM);
+ case FPOp::FRem:
+ return Result.mod(Args[1]);
+ case FPOp::MinNum:
+ Result = minnum(Result, Args[1]);
+ return getSNaNStatus(Args);
+ case FPOp::MaxNum:
+ Result = maxnum(Result, Args[1]);
+ return getSNaNStatus(Args);
+ case FPOp::Minimum:
+ Result = minimum(Result, Args[1]);
+ return getSNaNStatus(Args);
+ case FPOp::Maximum:
+ Result = maximum(Result, Args[1]);
+ return getSNaNStatus(Args);
+ case FPOp::MinimumNum:
+ Result = minimumnum(Result, Args[1]);
+ return getSNaNStatus(Args);
+ case FPOp::MaximumNum:
+ Result = maximumnum(Result, Args[1]);
+ return getSNaNStatus(Args);
+ }
+ } else if (Args.size() == 1) {
+ switch (Opcode) {
+ default:
+ llvm_unreachable("unsupported single-arg FP folding op");
+ case FPOp::Ceil:
+ return Result.roundToIntegral(RoundingMode::TowardPositive);
+ case FPOp::Floor:
+ return Result.roundToIntegral(RoundingMode::TowardNegative);
+ case FPOp::Trunc:
+ return Result.roundToIntegral(RoundingMode::TowardZero);
+ case FPOp::Round:
+ return Result.roundToIntegral(RoundingMode::NearestTiesToAway);
+ case FPOp::RoundEven:
+ return Result.roundToIntegral(RoundingMode::NearestTiesToEven);
+ }
+ } else if (Args.size() == 3 && Opcode == FPOp::FMA) {
+ return Result.fusedMultiplyAdd(Args[1], Args[2], RM);
+ }
+ llvm_unreachable("unsupported FP folding op for given number of inputs");
+}
+
+/// Internal helper to fold or reject if we cannot give correct results.
+/// If RM is necessary for the operation, it must be valid and non-dynamic.
+/// Dynamic denormal modes reject the fold if subnormal inputs are present or
+/// if the output is subnormal and may need flushing.
+static std::optional<FPFoldResult> tryFoldFPOpImpl(FPOp Opcode,
+ ArrayRef<APFloat> Args,
+ RoundingMode RM,
+ DenormalMode Denorms) {
+ assert(!Args.empty() && Args.size() <= 3);
+ assert(Denorms.isValid() && "invalid denormal mode");
+
+ SmallVector<APFloat, 3> FlushedArgs;
+ for (unsigned I = 0; I < Args.size(); ++I) {
+ assert(&Args[I].getSemantics() == &Args[0].getSemantics());
+ if (Denorms.Input == DenormalMode::Dynamic && Args[I].isDenormal())
+ return std::nullopt;
+ FlushedArgs.push_back(flush(Args[I], Denorms.Input));
+ }
+
+ APFloat Result = FlushedArgs[0];
+ APFloat::opStatus Status = executeFold(Opcode, Result, FlushedArgs, RM);
+
+ if (Denorms.Output == DenormalMode::Dynamic && Result.isDenormal())
+ return std::nullopt;
+ if (Denorms.Output != DenormalMode::IEEE &&
+ mayHaveBeenRoundedFromSubnormal(Result, Status, RM))
+ return std::nullopt;
+ return FPFoldResult{flush(std::move(Result), Denorms.Output), Status};
+}
+
+// Fold FPOps that do not require a rounding mode (or use the default RN)
+std::optional<FPFoldResult> llvm::tryFoldFP(FPOp Opcode, ArrayRef<APFloat> Args,
+ DenormalMode Denorms) {
+ assert(!Args.empty());
+ return tryFoldFPOpImpl(Opcode, Args, RoundingMode::NearestTiesToEven,
+ Denorms);
+}
+
+// Fold FPOps that do require a rounding mode. Handle the possibility
+// that the RM is dynamic (fold with RN and allow the fold
+// only if no rounding is required).
+std::optional<FPFoldResult> llvm::tryFoldFPWithRM(FPOp Opcode,
+ ArrayRef<APFloat> Args,
+ RoundingMode RM,
+ DenormalMode Denorms) {
+ assert(RM != RoundingMode::Invalid && "invalid rounding mode");
+ assert(!Args.empty());
+ RoundingMode EvalRM =
+ RM == RoundingMode::Dynamic ? RoundingMode::NearestTiesToEven : RM;
+ auto Res = tryFoldFPOpImpl(Opcode, Args, EvalRM, Denorms);
+ if (RM == RoundingMode::Dynamic && Res && (Res->Status & APFloat::opInexact))
+ return std::nullopt;
+ return Res;
+}
+
+std::optional<APFloat::cmpResult> llvm::tryFoldFCmp(const APFloat &LHS,
+ const APFloat &RHS,
+ DenormalMode Denorms) {
+ assert(Denorms.isValid() && "invalid denormal mode");
+ if (Denorms.Input == DenormalMode::Dynamic &&
+ (LHS.isDenormal() || RHS.isDenormal()))
+ return std::nullopt;
+
+ APFloat LHSTmp = flush(LHS, Denorms.Input);
+ APFloat RHSTmp = flush(RHS, Denorms.Input);
+ return LHSTmp.compare(RHSTmp);
+}
diff --git a/llvm/test/CodeGen/NVPTX/denormal-constant-fold.ll b/llvm/test/CodeGen/NVPTX/denormal-constant-fold.ll
index ea017cd10f5639..578938532d7e68 100644
--- a/llvm/test/CodeGen/NVPTX/denormal-constant-fold.ll
+++ b/llvm/test/CodeGen/NVPTX/denormal-constant-fold.ll
@@ -2,9 +2,9 @@
; RUN: llc -O0 -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx70 -verify-machineinstrs < %s | FileCheck %s
; Use -O0 to skip EarlyCSE and let the constant operations reach the DAG folders.
-; FIXME: Honor the function denormal environment in all these constant folds.
+; Constant folds honor the function's per-type input/output denormal modes.
-; FIXME: Flush the subnormal inputs so all seven FP results store +0 and
+; Flush the subnormal inputs so all seven FP results store +0 and
; fcmp stores true instead of false.
define void @arithmetic(ptr %fadd_res, ptr %fsub_res, ptr %fmul_res, ptr %fdiv_res, ptr %fmuladd_res, ptr %fma_res, ptr %fcmp_res, ptr %ceil_res) denormal_fpenv(float: preservesign) {
; CHECK-LABEL: arithmetic(
@@ -14,20 +14,20 @@ define void @arithmetic(ptr %fadd_res, ptr %fsub_res, ptr %fmul_res, ptr %fdiv_r
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b64 %rd1, [arithmetic_param_0];
; CHECK-NEXT: ld.param.b64 %rd2, [arithmetic_param_1];
-; CHECK-NEXT: st.b32 [%rd1], 2;
+; CHECK-NEXT: st.b32 [%rd1], 0;
; CHECK-NEXT: ld.param.b64 %rd3, [arithmetic_param_2];
; CHECK-NEXT: ld.param.b64 %rd4, [arithmetic_param_3];
-; CHECK-NEXT: st.b32 [%rd2], 2;
+; CHECK-NEXT: st.b32 [%rd2], 0;
; CHECK-NEXT: ld.param.b64 %rd5, [arithmetic_param_4];
; CHECK-NEXT: ld.param.b64 %rd6, [arithmetic_param_5];
-; CHECK-NEXT: st.b32 [%rd3], 2;
+; CHECK-NEXT: st.b32 [%rd3], 0;
; CHECK-NEXT: ld.param.b64 %rd7, [arithmetic_param_6];
; CHECK-NEXT: ld.param.b64 %rd8, [arithmetic_param_7];
-; CHECK-NEXT: st.b32 [%rd4], 2;
-; CHECK-NEXT: st.b32 [%rd5], 1;
-; CHECK-NEXT: st.b32 [%rd6], 1;
-; CHECK-NEXT: st.b8 [%rd7], 0;
-; CHECK-NEXT: st.b32 [%rd8], 1065353216;
+; CHECK-NEXT: st.b32 [%rd4], 0;
+; CHECK-NEXT: st.b32 [%rd5], 0;
+; CHECK-NEXT: st.b32 [%rd6], 0;
+; CHECK-NEXT: st.b8 [%rd7], 1;
+; CHECK-NEXT: st.b32 [%rd8], 0;
; CHECK-NEXT: ret;
%fadd = fadd float f0x00000001, f0x00000001
store float %fadd, ptr %fadd_res
@@ -48,7 +48,7 @@ define void @arithmetic(ptr %fadd_res, ptr %fsub_res, ptr %fmul_res, ptr %fdiv_r
ret void
}
-; FIXME: Flush the subnormal inputs so all six min/max results store +0.
+; Flush the subnormal inputs so all six min/max results store +0.
define void @minmax(ptr %minnum_res, ptr %maxnum_res, ptr %minimum_res, ptr %maximum_res, ptr %minimumnum_res, ptr %maximumnum_res) denormal_fpenv(float: preservesign) {
; CHECK-LABEL: minmax(
; CHECK: {
@@ -57,16 +57,16 @@ define void @minmax(ptr %minnum_res, ptr %maxnum_res, ptr %minimum_res, ptr %max
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b64 %rd1, [minmax_param_0];
; CHECK-NEXT: ld.param.b64 %rd2, [minmax_param_1];
-; CHECK-NEXT: st.b32 [%rd1], 1;
+; CHECK-NEXT: st.b32 [%rd1], 0;
; CHECK-NEXT: ld.param.b64 %rd3, [minmax_param_2];
; CHECK-NEXT: ld.param.b64 %rd4, [minmax_param_3];
-; CHECK-NEXT: st.b32 [%rd2], 1;
+; CHECK-NEXT: st.b32 [%rd2], 0;
; CHECK-NEXT: ld.param.b64 %rd5, [minmax_param_4];
; CHECK-NEXT: ld.param.b64 %rd6, [minmax_param_5];
-; CHECK-NEXT: st.b32 [%rd3], 1;
-; CHECK-NEXT: st.b32 [%rd4], 1;
-; CHECK-NEXT: st.b32 [%rd5], 1;
-; CHECK-NEXT: st.b32 [%rd6], 1;
+; CHECK-NEXT: st.b32 [%rd3], 0;
+; CHECK-NEXT: st.b32 [%rd4], 0;
+; CHECK-NEXT: st.b32 [%rd5], 0;
+; CHECK-NEXT: st.b32 [%rd6], 0;
; CHECK-NEXT: ret;
%minnum = call float @llvm.minnum.f32(float f0x00000001, float 1.000000e+00)
store float %minnum, ptr %minnum_res
@@ -83,26 +83,32 @@ define void @minmax(ptr %minnum_res, ptr %maxnum_res, ptr %minimum_res, ptr %max
ret void
}
-; FIXME: Leave these operations unfolded instead of storing minimum-normal.
+; Leave these operations unfolded instead of storing minimum-normal.
; Their exact results are subnormal and FTZ produces signed zero before rounding.
define void @ftz_boundaries(ptr %fmul_ftz_boundary_res, ptr %fdiv_ftz_boundary_res, ptr %fma_ftz_boundary_res, ptr %fmul_ftz_boundary_negative_res) denormal_fpenv(float: preservesign) {
; CHECK-LABEL: ftz_boundaries(
; CHECK: {
+; CHECK-NEXT: .reg .b32 %r<5>;
; CHECK-NEXT: .reg .b64 %rd<5>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b64 %rd1, [ftz_boundaries_param_0];
; CHECK-NEXT: ld.param.b64 %rd2, [ftz_boundaries_param_1];
-; CHECK-NEXT: st.b32 [%rd1], 8388608;
+; CHECK-NEXT: mul.rn.ftz.f32 %r1, 0f00800000, 0f3F7FFFFF;
+; CHECK-NEXT: st.b32 [%rd1], %r1;
; CHECK-NEXT: ld.param.b64 %rd3, [ftz_boundaries_param_2];
; CHECK-NEXT: ld.param.b64 %rd4, [ftz_boundaries_param_3];
-; CHECK-NEXT: st.b32 [%rd2], 8388608;
-; CHECK-NEXT: st.b32 [%rd3], 8388608;
-; CHECK-NEXT: st.b32 [%rd4], -2139095040;
+; CHECK-NEXT: mul.rn.ftz.f32 %r2, 0f00FFFFFF, 0f3F000000;
+; CHECK-NEXT: st.b32 [%rd2], %r2;
+; CHECK-NEXT: fma.rn.ftz.f32 %r3, 0f00800000, 0f3F7FFFFF, 0f00000000;
+; CHECK-NEXT: st.b32 [%rd3], %r3;
+; CHECK-NEXT: mul.rn.ftz.f32 %r4, 0f80800000, 0f3F7FFFFF;
+; CHECK-NEXT: st.b32 [%rd4], %r4;
; CHECK-NEXT: ret;
%fmul_ftz_boundary = fmul float f0x00800000, f0x3F7FFFFF
store float %fmul_ftz_boundary, ptr %fmul_ftz_boundary_res
- ; Division by two is folded using gradual underflow, as is multiplication.
+ ; Division by two may become multiplication by 0.5, but must not be
+ ; constant-folded using gradual underflow.
%fdiv_ftz_boundary = fdiv float f0x00FFFFFF, 2.000000e+00
store float %fdiv_ftz_boundary, ptr %fdiv_ftz_boundary_res
%fma_ftz_boundary = call float @llvm.fma.f32(float f0x00800000, float f0x3F7FFFFF, float 0.0)
@@ -112,27 +118,37 @@ define void @ftz_boundaries(ptr %fmul_ftz_boundary_res, ptr %fdiv_ftz_boundary_r
ret void
}
-; FIXME: Keep all six floating-point operations unfolded: their results depend
+; Keep all six floating-point operations unfolded: their results depend
; on dynamic input/output flushing. ceil can produce 1 or +0, and fcmp can
; produce false or true; the other results can be subnormal or +0.
define void @dynamic(ptr %fadd_res, ptr %fma_res, ptr %ceil_res, ptr %minnum_res, ptr %fcmp_res, ptr %output_mul_res) denormal_fpenv(float: dynamic) {
; CHECK-LABEL: dynamic(
; CHECK: {
+; CHECK-NEXT: .reg .pred %p<2>;
+; CHECK-NEXT: .reg .b16 %rs<2>;
+; CHECK-NEXT: .reg .b32 %r<6>;
; CHECK-NEXT: .reg .b64 %rd<7>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b64 %rd1, [dynamic_param_0];
; CHECK-NEXT: ld.param.b64 %rd2, [dynamic_param_1];
-; CHECK-NEXT: st.b32 [%rd1], 2;
+; CHECK-NEXT: add.rn.f32 %r1, 0f00000001, 0f00000001;
+; CHECK-NEXT: st.b32 [%rd1], %r1;
; CHECK-NEXT: ld.param.b64 %rd3, [dynamic_param_2];
; CHECK-NEXT: ld.param.b64 %rd4, [dynamic_param_3];
-; CHECK-NEXT: st.b32 [%rd2], 1;
+; CHECK-NEXT: add.rn.f32 %r2, 0f00000001, 0f00000000;
+; CHECK-NEXT: st.b32 [%rd2], %r2;
; CHECK-NEXT: ld.param.b64 %rd5, [dynamic_param_4];
; CHECK-NEXT: ld.param.b64 %rd6, [dynamic_param_5];
-; CHECK-NEXT: st.b32 [%rd3], 1065353216;
-; CHECK-NEXT: st.b32 [%rd4], 1;
-; CHECK-NEXT: st.b8 [%rd5], 0;
-; CHECK-NEXT: st.b32 [%rd6], 4194304;
+; CHECK-NEXT: cvt.rpi.f32.f32 %r3, 0f00000001;
+; CHECK-NEXT: st.b32 [%rd3], %r3;
+; CHECK-NEXT: min.f32 %r4, 0f00000001, 0f3F800000;
+; CHECK-NEXT: st.b32 [%rd4], %r4;
+; CHECK-NEXT: setp.eq.f32 %p1, 0f00000001, 0f00000000;
+; CHECK-NEXT: selp.b16 %rs1, 1, 0, %p1;
+; CHECK-NEXT: st.b8 [%rd5], %rs1;
+; CHECK-NEXT: mul.rn.f32 %r5, 0f00800000, 0f3F000000;
+; CHECK-NEXT: st.b32 [%rd6], %r5;
; CHECK-NEXT: ret;
%fadd = fadd float f0x00000001, f0x00000001
store float %fadd, ptr %fadd_res
@@ -151,7 +167,7 @@ define void @dynamic(ptr %fadd_res, ptr %fma_res, ptr %ceil_res, ptr %minnum_res
; Each f16 result below is stored in its own eight-byte slot.
-; FIXME: All FP results should store +0, and fcmp should store true (zext to 1),
+; All FP results should store +0, and fcmp should store true (zext to 1),
; independently of float: ieee. %output_mul has normal inputs but a subnormal
; result, so it tests output flushing separately from input flushing.
define void @f16_ftz(ptr %out) denormal_fpenv(preservesign, float: ieee) {
@@ -161,20 +177,20 @@ define void @f16_ftz(ptr %out) denormal_fpenv(preservesign, float: ieee) {
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b64 %rd1, [f16_ftz_param_0];
-; CHECK-NEXT: st.b16 [%rd1], 0x0002;
-; CHECK-NEXT: st.b16 [%rd1+8], 0x0002;
-; CHECK-NEXT: st.b16 [%rd1+16], 0x0002;
-; CHECK-NEXT: st.b16 [%rd1+24], 0x0002;
-; CHECK-NEXT: st.b16 [%rd1+32], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+40], 0x3C00;
-; CHECK-NEXT: st.b16 [%rd1+48], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+56], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+64], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+72], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+80], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+88], 0x0001;
-; CHECK-NEXT: st.b32 [%rd1+96], 0;
-; CHECK-NEXT: st.b16 [%rd1+104], 0x0200;
+; CHECK-NEXT: st.b16 [%rd1], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+8], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+16], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+24], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+32], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+40], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+48], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+56], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+64], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+72], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+80], 0x0000;
+; CHECK-NEXT: st.b16 [%rd1+88], 0x0000;
+; CHECK-NEXT: st.b32 [%rd1+96], 1;
+; CHECK-NEXT: st.b16 [%rd1+104], 0x0000;
; CHECK-NEXT: ret;
%fadd = fadd half f0x0001, f0x0001
%fadd_res = getelementptr i8, ptr %out, i64 0
@@ -222,22 +238,37 @@ define void @f16_ftz(ptr %out) denormal_fpenv(preservesign, float: ieee) {
ret void
}
-; FIXME: Keep all six floating-point operations unfolded: the default dynamic
+; Keep all six floating-point operations unfolded: the default dynamic
; mode applies to f16 despite float: ieee. ceil can produce 1 or +0, fcmp
; can produce false or true, and the other results can be subnormal or +0.
define void @f16_dynamic(ptr %out) denormal_fpenv(dynamic, float: ieee) {
; CHECK-LABEL: f16_dynamic(
; CHECK: {
+; CHECK-NEXT: .reg .pred %p<2>;
+; CHECK-NEXT: .reg .b16 %rs<11>;
+; CHECK-NEXT: .reg .b32 %r<2>;
; CHECK-NEXT: .reg .b64 %rd<2>;
; CHECK-EMPTY:
; CHECK-NEXT: // %bb.0:
; CHECK-NEXT: ld.param.b64 %rd1, [f16_dynamic_param_0];
-; CHECK-NEXT: st.b16 [%rd1], 0x0002;
-; CHECK-NEXT: st.b16 [%rd1+8], 0x0001;
-; CHECK-NEXT: st.b16 [%rd1+16], 0x3C00;
-; CHECK-NEXT: st.b16 [%rd1+24], 0x0001;
-; CHECK-NEXT: st.b32 [%rd1+32], 0;
-; CHECK-NEXT: st.b16 [%rd1+40], 0x0200;
+; CHECK-NEXT: mov.b16 %rs1, 0x0001;
+; CHECK-NEXT: add.rn.f16 %rs2, %rs1, %rs1;
+; CHECK-NEXT: st.b16 [%rd1], %rs2;
+; CHECK-NEXT: mov.b16 %rs3, 0x0000;
+; CHECK-NEXT: add.rn.f16 %rs4, %rs1, %rs3;
+; CHECK-NEXT: st.b16 [%rd1+8], %rs4;
+; CHECK-NEXT: cvt.rpi.f16.f16 %rs5, %rs1;
+; CHECK-NEXT: st.b16 [%rd1+16], %rs5;
+; CHECK-NEXT: mov.b16 %rs6, 0x3C00;
+; CHECK-NEXT: min.f16 %rs7, %rs1, %rs6;
+; CHECK-NEXT: st.b16 [%rd1+24], %rs7;
+; CHECK-NEXT: setp.eq.f16 %p1, %rs1, %rs3;
+; CHECK-NEXT: selp.b32 %r1, 1, 0, %p1;
+; CHECK-NEXT: st.b32 [%rd1+32], %r1;
+; CHECK-NEXT: mov.b16 %rs8, 0x3800;
+; CHECK-NEXT: mov.b16 %rs9, 0x0400;
+; CHECK-NEXT: mul.rn.f16 %rs10, %rs9, %rs8;
+; CHECK-NEXT: st.b16 [%rd1+40], %rs10;
; CHECK-NEXT: ret;
%fadd = fadd half f0x0001, f0x0001
%fadd_res = getelementptr i8, ptr %out, i64 0
diff --git a/llvm/utils/gn/secondary/llvm/lib/Support/BUILD.gn b/llvm/utils/gn/secondary/llvm/lib/Support/BUILD.gn
index 35f364e073a69c..866b87b7e0ef66 100644
--- a/llvm/utils/gn/secondary/llvm/lib/Support/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/lib/Support/BUILD.gn
@@ -96,6 +96,7 @@ static_library("Support") {
"FileOutputBuffer.cpp",
"FileUtilities.cpp",
"FloatingPointMode.cpp",
+ "FPFold.cpp",
"FoldingSet.cpp",
"FormatVariadic.cpp",
"FormattedStream.cpp",
>From 57e6c033bbf98e0b9e6380f8fc6b05d846ef2f2b Mon Sep 17 00:00:00 2001
From: Tim Gymnich <tim at gymni.ch>
Date: Sat, 21 Nov 2026 13:04:03 +0100
Subject: [PATCH 15/35] [ValueTracking] Track constant ExtractElement indices
in sign-bit queries (#244296)
Query only the selected lane for an in-range constant index into a fixed
vector.
This matches the behavior in SDAG.
Assisted by: GPT 6.1 Sol
---
llvm/lib/Analysis/ValueTracking.cpp | 19 +++++--
.../ValueTracking/signbits-extract-elt.ll | 56 +++++++++++++++++++
2 files changed, 69 insertions(+), 6 deletions(-)
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index d781fb9591c4a4..65881dec7f1ab4 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -4895,12 +4895,19 @@ static unsigned ComputeNumSignBitsImpl(const Value *V,
return 1;
}
- case Instruction::ExtractElement:
- // Look through extract element. At the moment we keep this simple and
- // skip tracking the specific element. But at least we might find
- // information valid for all elements of the vector (for example if vector
- // is sign extended, shifted, etc).
- return ComputeNumSignBits(U->getOperand(0), Q, Depth + 1);
+ case Instruction::ExtractElement: {
+ const Value *Vec = U->getOperand(0);
+ auto *VecTy = dyn_cast<FixedVectorType>(Vec->getType());
+ if (!VecTy)
+ return ComputeNumSignBits(Vec, Q, Depth + 1);
+
+ unsigned NumElts = VecTy->getNumElements();
+ APInt DemandedVecElts = APInt::getAllOnes(NumElts);
+ if (auto *CIdx = dyn_cast<ConstantInt>(U->getOperand(1)))
+ if (CIdx->getValue().ult(NumElts))
+ DemandedVecElts = APInt::getOneBitSet(NumElts, CIdx->getZExtValue());
+ return ComputeNumSignBits(Vec, DemandedVecElts, Q, Depth + 1);
+ }
case Instruction::ShuffleVector: {
// Collect the minimum number of sign bits that are shared by every vector
diff --git a/llvm/test/Analysis/ValueTracking/signbits-extract-elt.ll b/llvm/test/Analysis/ValueTracking/signbits-extract-elt.ll
index f74f025611cfcf..338f9ecfdb323f 100644
--- a/llvm/test/Analysis/ValueTracking/signbits-extract-elt.ll
+++ b/llvm/test/Analysis/ValueTracking/signbits-extract-elt.ll
@@ -121,3 +121,59 @@ define <2 x i32> @bitcast_scalar_sign_splat(i1 %x) {
%result = ashr <2 x i32> %cast, splat (i32 5)
ret <2 x i32> %result
}
+
+define i32 @extract_demanded(<2 x i1> %x, <2 x i32> %wide) {
+; CHECK-LABEL: @extract_demanded(
+; CHECK-NEXT: [[EXT:%.*]] = sext <2 x i1> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT: [[SRC:%.*]] = shufflevector <2 x i32> [[EXT]], <2 x i32> [[WIDE:%.*]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT: [[ELT:%.*]] = extractelement <2 x i32> [[SRC]], i32 0
+; CHECK-NEXT: ret i32 [[ELT]]
+;
+ %ext = sext <2 x i1> %x to <2 x i32>
+ %src = shufflevector <2 x i32> %ext, <2 x i32> %wide, <2 x i32> <i32 0, i32 2>
+ %elt = extractelement <2 x i32> %src, i32 0
+ %result = ashr i32 %elt, 5
+ ret i32 %result
+}
+
+define i32 @extract_other(<2 x i1> %x, <2 x i32> %wide) {
+; CHECK-LABEL: @extract_other(
+; CHECK-NEXT: [[EXT:%.*]] = sext <2 x i1> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT: [[SRC:%.*]] = shufflevector <2 x i32> [[EXT]], <2 x i32> [[WIDE:%.*]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT: [[ELT:%.*]] = extractelement <2 x i32> [[SRC]], i32 1
+; CHECK-NEXT: [[RESULT:%.*]] = ashr i32 [[ELT]], 5
+; CHECK-NEXT: ret i32 [[RESULT]]
+;
+ %ext = sext <2 x i1> %x to <2 x i32>
+ %src = shufflevector <2 x i32> %ext, <2 x i32> %wide, <2 x i32> <i32 0, i32 2>
+ %elt = extractelement <2 x i32> %src, i32 1
+ %result = ashr i32 %elt, 5
+ ret i32 %result
+}
+
+define i32 @extract_variable(<2 x i1> %x, <2 x i32> %wide, i32 %idx) {
+; CHECK-LABEL: @extract_variable(
+; CHECK-NEXT: [[EXT:%.*]] = sext <2 x i1> [[X:%.*]] to <2 x i32>
+; CHECK-NEXT: [[SRC:%.*]] = shufflevector <2 x i32> [[EXT]], <2 x i32> [[WIDE:%.*]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT: [[ELT:%.*]] = extractelement <2 x i32> [[SRC]], i32 [[IDX:%.*]]
+; CHECK-NEXT: [[RESULT:%.*]] = ashr i32 [[ELT]], 5
+; CHECK-NEXT: ret i32 [[RESULT]]
+;
+ %ext = sext <2 x i1> %x to <2 x i32>
+ %src = shufflevector <2 x i32> %ext, <2 x i32> %wide, <2 x i32> <i32 0, i32 2>
+ %elt = extractelement <2 x i32> %src, i32 %idx
+ %result = ashr i32 %elt, 5
+ ret i32 %result
+}
+
+define i32 @extract_scalable(<vscale x 2 x i1> %x) {
+; CHECK-LABEL: @extract_scalable(
+; CHECK-NEXT: [[SRC:%.*]] = sext <vscale x 2 x i1> [[X:%.*]] to <vscale x 2 x i32>
+; CHECK-NEXT: [[ELT:%.*]] = extractelement <vscale x 2 x i32> [[SRC]], i32 0
+; CHECK-NEXT: ret i32 [[ELT]]
+;
+ %src = sext <vscale x 2 x i1> %x to <vscale x 2 x i32>
+ %elt = extractelement <vscale x 2 x i32> %src, i32 0
+ %result = ashr i32 %elt, 5
+ ret i32 %result
+}
>From 1d80abf1662cdbf76bd479176f5a04e40d94f909 Mon Sep 17 00:00:00 2001
From: Feng Zou <feng.zou at intel.com>
Date: Sat, 21 Nov 2026 13:05:24 +0100
Subject: [PATCH 16/35] [InlineSpiller] Do not merge or hoist a spill with no
original value (#237148)
HoistSpillHelper keeps a copy of the original live interval, taken at
the first spill to each stack slot and never updated, and keys
MergeableSpills by the original VNInfo of each spill. Since #199182,
foldMemoryOperand can extend the original interval backwards over the
COPY that an X86 NDD fold inserts (#191368). A spill later folded into
that COPY lies outside the saved interval, so getVNInfoAt returns null.
The null VNInfo* is stored as a MergeableSpills key and dereferenced in
runHoistSpills (SIGSEGV).
Leave such a spill out of merging and hoisting; it stays where it was
inserted. Both are optimizations, so skipping them is safe, and inputs
that did not crash before get the same code.
Assisted-by: Claude Opus 5.5 <noreply at anthropic.com>
---
llvm/lib/CodeGen/InlineSpiller.cpp | 7 +
.../apx/inline-spiller-hoist-folded-copy.ll | 259 ++++++++++++++++++
2 files changed, 266 insertions(+)
create mode 100644 llvm/test/CodeGen/X86/apx/inline-spiller-hoist-folded-copy.ll
diff --git a/llvm/lib/CodeGen/InlineSpiller.cpp b/llvm/lib/CodeGen/InlineSpiller.cpp
index 063940a6bfe4af..06b279c5301d92 100644
--- a/llvm/lib/CodeGen/InlineSpiller.cpp
+++ b/llvm/lib/CodeGen/InlineSpiller.cpp
@@ -1499,6 +1499,13 @@ void HoistSpillHelper::addToMergeableSpills(MachineInstr &Spill, int StackSlot,
SlotIndex Idx = LIS.getInstructionIndex(Spill);
VNInfo *OrigVNI = Place->second->getVNInfoAt(Idx.getRegSlot());
+ // OrigVNI is null if InlineSpiller::foldMemoryOperand() extended Original's
+ // interval after it was saved above: a fold that inserts a COPY (e.g. X86
+ // NDD to two-address form) makes Original live from that COPY, and a spill
+ // later folded into the COPY lands there. That spill stores the COPY's
+ // source, not the folded instruction's result, so don't merge or hoist it.
+ if (!OrigVNI)
+ return;
std::pair<int, VNInfo *> MIdx = std::make_pair(StackSlot, OrigVNI);
MergeableSpills[MIdx].insert(&Spill);
}
diff --git a/llvm/test/CodeGen/X86/apx/inline-spiller-hoist-folded-copy.ll b/llvm/test/CodeGen/X86/apx/inline-spiller-hoist-folded-copy.ll
new file mode 100644
index 00000000000000..2d41fd4a6af090
--- /dev/null
+++ b/llvm/test/CodeGen/X86/apx/inline-spiller-hoist-folded-copy.ll
@@ -0,0 +1,259 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+ndd < %s | FileCheck %s
+
+; With NDD but not prefer-ndd-mem, folding a reload into an NDD instruction
+; emits a COPY of the other source plus the two-address form, and InlineSpiller
+; extends the original live interval back to the COPY. A spill later folded into
+; that COPY is outside HoistSpillHelper's earlier snapshot of the interval,
+; which used to cause a null VNInfo dereference.
+
+declare i64 @llvm.smax.i64(i64, i64)
+
+define void @test(ptr %arg, ptr %arg1, ptr %arg2, ptr %arg3, ptr %arg4, i1 %arg5, i64 %arg6, i64 %arg7, i1 %arg8, i64 %arg9) nounwind {
+; CHECK-LABEL: test:
+; CHECK: # %bb.0: # %bb
+; CHECK-NEXT: pushq %rbp
+; CHECK-NEXT: pushq %r15
+; CHECK-NEXT: pushq %r14
+; CHECK-NEXT: pushq %r13
+; CHECK-NEXT: pushq %r12
+; CHECK-NEXT: pushq %rbx
+; CHECK-NEXT: subq $32, %rsp
+; CHECK-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rax
+; CHECK-NEXT: movzbl {{[0-9]+}}(%rsp), %r10d
+; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rbx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: leaq (%rcx,%rax,8), %rcx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: shlq $3, %rbx, %rcx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: shlq $2, %rbx, %rcx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: xorps %xmm0, %xmm0
+; CHECK-NEXT: movl %r9d, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
+; CHECK-NEXT: .p2align 4
+; CHECK-NEXT: .LBB0_1: # %bb10
+; CHECK-NEXT: # =>This Loop Header: Depth=1
+; CHECK-NEXT: # Child Loop BB0_2 Depth 2
+; CHECK-NEXT: # Child Loop BB0_4 Depth 2
+; CHECK-NEXT: # Child Loop BB0_5 Depth 3
+; CHECK-NEXT: # Child Loop BB0_8 Depth 4
+; CHECK-NEXT: movslq 0, %rdx
+; CHECK-NEXT: movq 0, %rsi
+; CHECK-NEXT: .p2align 4
+; CHECK-NEXT: .LBB0_2: # %bb13
+; CHECK-NEXT: # Parent Loop BB0_1 Depth=1
+; CHECK-NEXT: # => This Inner Loop Header: Depth=2
+; CHECK-NEXT: testb $1, %r9b
+; CHECK-NEXT: je .LBB0_2
+; CHECK-NEXT: # %bb.3: # %bb14
+; CHECK-NEXT: # in Loop: Header=BB0_1 Depth=1
+; CHECK-NEXT: movq 0, %rbp
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; CHECK-NEXT: movq (%rcx), %r14
+; CHECK-NEXT: movl %edx, %edi
+; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %rcx
+; CHECK-NEXT: testq %rcx, %rcx
+; CHECK-NEXT: movl $0, %r8d
+; CHECK-NEXT: cmovleq %r8, %rcx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r8 # 8-byte Reload
+; CHECK-NEXT: imulq %rdi, %r8, %r12
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Folded Reload
+; CHECK-NEXT: shlq $2, %rsi, %r15
+; CHECK-NEXT: shlq $2, %rdi, %r11
+; CHECK-NEXT: leaq (%r11,%rsi,4), %rsi
+; CHECK-NEXT: imulq %rdx, %rsi, %r8
+; CHECK-NEXT: shlq $2, %rdx, %rcx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: shlq $3, %rdx, %rbx
+; CHECK-NEXT: shlq $3, %rdi, %rsi
+; CHECK-NEXT: leaq (%rsi,%rdx,8), %rcx
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; CHECK-NEXT: movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: imulq %rdi, %rbx
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Folded Reload
+; CHECK-NEXT: movq %rcx, %rsi
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload
+; CHECK-NEXT: leaq (%rcx,%rdx,8), %rcx
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %rdi, %rcx
+; CHECK-NEXT: imulq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Folded Reload
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: leaq (%rcx,%rdx,4), %rdx
+; CHECK-NEXT: imulq %rdi, %rdx
+; CHECK-NEXT: movabsq $9223372036854775800, %rcx # imm = 0x7FFFFFFFFFFFFFF8
+; CHECK-NEXT: movq %rcx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Reload
+; CHECK-NEXT: andq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Folded Reload
+; CHECK-NEXT: movq %r13, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movl $1, %ecx
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r13 # 8-byte Reload
+; CHECK-NEXT: subq %r13, %rcx
+; CHECK-NEXT: movq {{[0-9]+}}(%rsp), %r13
+; CHECK-NEXT: jmp .LBB0_4
+; CHECK-NEXT: .p2align 4
+; CHECK-NEXT: .LBB0_10: # %bb65
+; CHECK-NEXT: # in Loop: Header=BB0_4 Depth=2
+; CHECK-NEXT: incq %rdi
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Reload
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %r12 # 8-byte Folded Reload
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r8 # 8-byte Reload
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %r8 # 8-byte Folded Reload
+; CHECK-NEXT: movq (%rsp), %rsi # 8-byte Reload
+; CHECK-NEXT: addq $8, %rsi
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Folded Reload
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; CHECK-NEXT: addq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Folded Reload
+; CHECK-NEXT: movl {{[-0-9]+}}(%r{{[sb]}}p), %r9d # 4-byte Reload
+; CHECK-NEXT: testb $1, %r9b
+; CHECK-NEXT: je .LBB0_1
+; CHECK-NEXT: .LBB0_4: # %bb22
+; CHECK-NEXT: # Parent Loop BB0_1 Depth=1
+; CHECK-NEXT: # => This Loop Header: Depth=2
+; CHECK-NEXT: # Child Loop BB0_5 Depth 3
+; CHECK-NEXT: # Child Loop BB0_8 Depth 4
+; CHECK-NEXT: movq %rbx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %r12, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %rdx, %r9
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
+; CHECK-NEXT: movq %rsi, (%rsp) # 8-byte Spill
+; CHECK-NEXT: movq %rsi, %r12
+; CHECK-NEXT: movq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
+; CHECK-NEXT: movq %r8, %rsi
+; CHECK-NEXT: jmp .LBB0_5
+; CHECK-NEXT: .p2align 4
+; CHECK-NEXT: .LBB0_9: # %bb63
+; CHECK-NEXT: # in Loop: Header=BB0_5 Depth=3
+; CHECK-NEXT: addq %r15, %rsi
+; CHECK-NEXT: addq $8, %r12
+; CHECK-NEXT: addq $8, %rdx
+; CHECK-NEXT: addq %r11, %r9
+; CHECK-NEXT: testb $1, %r10b
+; CHECK-NEXT: je .LBB0_10
+; CHECK-NEXT: .LBB0_5: # %bb24
+; CHECK-NEXT: # Parent Loop BB0_1 Depth=1
+; CHECK-NEXT: # Parent Loop BB0_4 Depth=2
+; CHECK-NEXT: # => This Loop Header: Depth=3
+; CHECK-NEXT: # Child Loop BB0_8 Depth 4
+; CHECK-NEXT: testb $1, %r10b
+; CHECK-NEXT: movq %r13, %rbx
+; CHECK-NEXT: jne .LBB0_7
+; CHECK-NEXT: # %bb.6: # %bb30
+; CHECK-NEXT: # in Loop: Header=BB0_5 Depth=3
+; CHECK-NEXT: movl $1, %r8d
+; CHECK-NEXT: imulq %r13, %r8
+; CHECK-NEXT: imulq %rdi, %r8
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rbx # 8-byte Reload
+; CHECK-NEXT: leaq (%rbx,%r8,8), %r8
+; CHECK-NEXT: movups %xmm0, 48(%r8,%rax,8)
+; CHECK-NEXT: movups %xmm0, 32(%r8,%rax,8)
+; CHECK-NEXT: movups %xmm0, 16(%r8,%rax,8)
+; CHECK-NEXT: movups %xmm0, (%r8,%rax,8)
+; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %r8 # 8-byte Reload
+; CHECK-NEXT: cmpq %r8, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Reload
+; CHECK-NEXT: movq %rcx, %rbx
+; CHECK-NEXT: je .LBB0_9
+; CHECK-NEXT: .LBB0_7: # %bb32.preheader
+; CHECK-NEXT: # in Loop: Header=BB0_5 Depth=3
+; CHECK-NEXT: xorl %r8d, %r8d
+; CHECK-NEXT: .p2align 4
+; CHECK-NEXT: .LBB0_8: # %bb32
+; CHECK-NEXT: # Parent Loop BB0_1 Depth=1
+; CHECK-NEXT: # Parent Loop BB0_4 Depth=2
+; CHECK-NEXT: # Parent Loop BB0_5 Depth=3
+; CHECK-NEXT: # => This Inner Loop Header: Depth=4
+; CHECK-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
+; CHECK-NEXT: movsd %xmm1, (%r12,%r8,2)
+; CHECK-NEXT: addq $4, %r8
+; CHECK-NEXT: testq %rbx, %rbx
+; CHECK-NEXT: movl $0, %ebx
+; CHECK-NEXT: jg .LBB0_8
+; CHECK-NEXT: jmp .LBB0_9
+bb:
+ br label %bb10
+
+bb10: ; preds = %bb65, %bb
+ %i = load i32, ptr null, align 4
+ %i11 = load i64, ptr null, align 8
+ %i12 = sext i32 %i to i64
+ br label %bb13
+
+bb13: ; preds = %bb13, %bb10
+ br i1 %arg5, label %bb14, label %bb13
+
+bb14: ; preds = %bb13
+ %i15 = load ptr, ptr null, align 8
+ %i16 = load i64, ptr %arg, align 8
+ %i17 = zext i32 %i to i64
+ %i18 = add i64 0, 1
+ %i19 = call i64 @llvm.smax.i64(i64 %arg7, i64 0)
+ %i20 = and i64 %i19, -8
+ %i21 = sub i64 1, %i20
+ br label %bb22
+
+bb22: ; preds = %bb65, %bb14
+ %i23 = phi i64 [ %i17, %bb14 ], [ %i66, %bb65 ]
+ br label %bb24
+
+bb24: ; preds = %bb63, %bb22
+ %i25 = phi i64 [ %i12, %bb22 ], [ %i64, %bb63 ]
+ %i26 = mul i64 %i18, %arg6
+ %i27 = mul i64 %i26, %i23
+ %i28 = getelementptr [8 x i8], ptr %arg3, i64 %i27
+ %i29 = getelementptr [8 x i8], ptr %i28, i64 %arg9
+ br i1 %arg8, label %bb32, label %bb30
+
+bb30: ; preds = %bb24
+ store <8 x double> zeroinitializer, ptr %i29, align 8
+ %i31 = icmp eq i64 %i19, %i20
+ br i1 %i31, label %bb63, label %bb32
+
+bb32: ; preds = %bb32, %bb30, %bb24
+ %i33 = phi i64 [ %i61, %bb32 ], [ 0, %bb30 ], [ 0, %bb24 ]
+ %i34 = phi i64 [ 0, %bb32 ], [ %i21, %bb30 ], [ %arg6, %bb24 ]
+ %i35 = mul i64 %arg6, %i17
+ %i36 = mul i64 %i35, %i23
+ %i37 = getelementptr [4 x i8], ptr null, i64 %i36
+ %i38 = mul i64 %i25, %i17
+ %i39 = getelementptr [4 x i8], ptr %i37, i64 %i38
+ %i40 = getelementptr [4 x i8], ptr %i39, i64 %i33
+ %i41 = load i32, ptr %i40, align 4
+ %i42 = getelementptr [8 x i8], ptr %arg1, i64 %i25
+ %i43 = getelementptr [8 x i8], ptr %i42, i64 %i33
+ %i44 = load double, ptr %i43, align 8
+ %i45 = mul i64 %i23, %i12
+ %i46 = getelementptr [8 x i8], ptr %arg4, i64 %i45
+ %i47 = getelementptr [8 x i8], ptr %i46, i64 %i33
+ %i48 = load double, ptr %i47, align 8
+ %i49 = getelementptr [8 x i8], ptr %i15, i64 %i16
+ %i50 = load double, ptr %i49, align 8
+ %i51 = getelementptr [8 x i8], ptr %arg2, i64 %i23
+ %i52 = getelementptr [8 x i8], ptr %i51, i64 %i25
+ %i53 = getelementptr [8 x i8], ptr %i52, i64 %i33
+ store double %i50, ptr %i53, align 8
+ %i54 = getelementptr [4 x i8], ptr null, i64 %i45
+ %i55 = mul i64 %i25, %i11
+ %i56 = getelementptr [4 x i8], ptr %i54, i64 %i55
+ %i57 = getelementptr [4 x i8], ptr %i56, i64 %i33
+ %i58 = load i32, ptr %i57, align 4
+ %i59 = getelementptr [8 x i8], ptr %i29, i64 %i33
+ %i60 = load double, ptr %i59, align 8
+ %i61 = add i64 %i33, 1
+ %i62 = icmp sgt i64 %i34, 0
+ br i1 %i62, label %bb32, label %bb63
+
+bb63: ; preds = %bb32, %bb30
+ %i64 = add i64 %i25, 1
+ br i1 %arg8, label %bb24, label %bb65
+
+bb65: ; preds = %bb63
+ %i66 = add i64 %i23, 1
+ br i1 %arg5, label %bb22, label %bb10
+}
>From 7016ba10ad070658b9c62f47b8fde16be7f2ac4c Mon Sep 17 00:00:00 2001
From: Feng Zou <feng.zou at intel.com>
Date: Sat, 21 Nov 2026 13:05:52 +0100
Subject: [PATCH 17/35] [InlineSpiller] Use a new value number for the COPY
range of a fold (#243055)
When folding a memory operand makes the target insert a COPY (e.g. a
reload folded into an X86 NDD instruction), foldMemoryOperand() extends
the original live interval from the COPY to the folded instruction
(#199182). That segment reused the folded instruction's value number,
though the register holds the COPY's source there.
HoistSpillHelper groups spills by (stack slot, original value number),
so rmRedundantSpills treated the store the COPY was folded into and the
later spill of the folded result as duplicates, and deleted the latter.
In memfold-origVNI-crash.ll this dropped an `or`, returning %p2 instead
of %p2 | %p3.
Give the segment its own value number at the COPY. Replace the
getVNInfoAt() guard with an assert: the sub-interval invariant already
requires OrigLI to be live at FoldMI's def.
Assisted-by: Claude Opus 5.5 <noreply at anthropic.com>
---
llvm/lib/CodeGen/InlineSpiller.cpp | 14 ++++++++++----
llvm/test/CodeGen/X86/apx/memfold-origVNI-crash.ll | 7 +++++++
2 files changed, 17 insertions(+), 4 deletions(-)
diff --git a/llvm/lib/CodeGen/InlineSpiller.cpp b/llvm/lib/CodeGen/InlineSpiller.cpp
index 06b279c5301d92..5e69cd4f72f019 100644
--- a/llvm/lib/CodeGen/InlineSpiller.cpp
+++ b/llvm/lib/CodeGen/InlineSpiller.cpp
@@ -1151,12 +1151,18 @@ foldMemoryOperand(ArrayRef<std::pair<MachineInstr *, unsigned>> Ops,
// sub-interval invariant holds: the original must be live wherever
// any of its children are live. Without this, reMaterializeFor()
// can query OrigLI at an early-clobber slot that falls inside
- // [CopyIdx, FoldIdx) and get a null VNI, triggering an assertion.
+ // [CopyIdx, FoldIdx.getRegSlot()) and get a null VNI, triggering an
+ // assertion. Use a new value number: CopyDstReg holds the COPY's
+ // source there, not FoldMI's result, and HoistSpillHelper merges
+ // spills that share a value number.
assert(LIS.hasInterval(OrigReg) && "OrigReg should have live interval");
LiveInterval &OrigLI = LIS.getInterval(OrigReg);
- if (VNInfo *OrigVNI = OrigLI.getVNInfoAt(FoldIdx.getRegSlot()))
- OrigLI.addSegment(
- LiveRange::Segment(CopyIdx, FoldIdx.getRegSlot(), OrigVNI));
+ assert(OrigLI.liveAt(FoldIdx.getRegSlot()) &&
+ "OrigLI should be live at FoldMI's def");
+ VNInfo *CopyVNI =
+ OrigLI.getNextValue(CopyIdx, LIS.getVNInfoAllocator());
+ OrigLI.addSegment(
+ LiveRange::Segment(CopyIdx, FoldIdx.getRegSlot(), CopyVNI));
}
}
}
diff --git a/llvm/test/CodeGen/X86/apx/memfold-origVNI-crash.ll b/llvm/test/CodeGen/X86/apx/memfold-origVNI-crash.ll
index 283e7b3454f6c3..a3e8c99a71361e 100644
--- a/llvm/test/CodeGen/X86/apx/memfold-origVNI-crash.ll
+++ b/llvm/test/CodeGen/X86/apx/memfold-origVNI-crash.ll
@@ -1,6 +1,10 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
; RUN: llc < %s -mtriple=x86_64-- -mattr=ndd -verify-machineinstrs | FileCheck %s
+; Folding a reload into the NDD `or` that defines %v3 inserts a COPY. The
+; COPY's range in the original live interval needs its own value number, or
+; HoistSpillHelper drops the spill of %v3 and @foo returns %p2.
+
define fastcc i64 @foo(ptr %p0, ptr %p1, i64 %p2, i64 %p3, ptr %p4, ptr %p5, ptr %p6, ptr %p7, ptr %p8, i64 %p9, i1 %cond, i32 %n) nounwind {
; CHECK-LABEL: foo:
; CHECK: # %bb.0: # %entry
@@ -110,6 +114,9 @@ define fastcc i64 @foo(ptr %p0, ptr %p1, i64 %p2, i64 %p3, ptr %p4, ptr %p5, ptr
; CHECK-NEXT: movq %rbx, %rdx
; CHECK-NEXT: callq memcpy at PLT
; CHECK-NEXT: movq %rbx, (%rsp) # 8-byte Spill
+; CHECK-NEXT: movq (%rsp), %rax # 8-byte Reload
+; CHECK-NEXT: orq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Folded Reload
+; CHECK-NEXT: movq %rax, (%rsp) # 8-byte Spill
; CHECK-NEXT: .LBB0_27: # %exit
; CHECK-NEXT: movq (%rsp), %rax # 8-byte Reload
; CHECK-NEXT: jmp .LBB0_8
>From 6e57e212b9d181388c5d6ae0b06db9b4f457b1a6 Mon Sep 17 00:00:00 2001
From: Lewis Crawford <lcrawford at nvidia.com>
Date: Sat, 21 Nov 2026 12:10:25 +0000
Subject: [PATCH 18/35] [AArch64][GlobalISel][NFC] Add denormal const-fold
baselines (#241382)
Add GlobalISel lit tests for AArch64 recording the current scalar
constant folding behaviour for f16, f32 and f64 under IEEE,
flush-to-zero and dynamic denormal environments.
Add TODO comments describing the desired results if const-folding were
to honour denormal_fpenv, which will be added in a follow-up commit.
---
.../GlobalISel/denormal-constant-fold.mir | 503 ++++++++++++++++++
1 file changed, 503 insertions(+)
create mode 100644 llvm/test/CodeGen/AArch64/GlobalISel/denormal-constant-fold.mir
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/denormal-constant-fold.mir b/llvm/test/CodeGen/AArch64/GlobalISel/denormal-constant-fold.mir
new file mode 100644
index 00000000000000..940a5817a423b2
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/denormal-constant-fold.mir
@@ -0,0 +1,503 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 6
+# RUN: llc -mtriple=aarch64 -mattr=+fullfp16 -run-pass=aarch64-prelegalizer-combiner \
+# RUN: -verify-machineinstrs %s -o - | FileCheck %s
+
+# Record the current generic GlobalISel folding behavior for arithmetic,
+# G_FMA, G_FCMP, rounding, and min/max. These folds currently ignore the
+# MachineFunction's denormal_fpenv attribute, even when set to "dynamic".
+#
+# Ensure the tests exercise combinations like:
+# denormal_fpenv(ieee, float: preservesign)
+# denormal_fpenv(preservesign, float:ieee)
+# to cover cases where f16 should follow the default `denormal_fpenv` but
+# f32 is specified differently.
+#
+# TODO: Make these constant folds honour the denormal environment.
+# The current values produced via constant folding are technically legal since
+# denormal_fpenv "may" flush but is not required to. However, it may be desirable in
+# most cases to have the const-folded results match the results that would be produced
+# if they exectuted on the actual hardware with non-const inputs containing the same values.
+#
+# It is not always possible to 100% match the const-folding to the hardware-executed result.
+# For example with f16 flushing enabled, f32 flushing disabled with:
+# denormal_fpenv(preservesign, float:ieee)
+# The input:
+# fadd half f0x0001, f0x0002
+# could be lowered to:
+# fadd h0, h0, h1
+# which would flush if FZ16 was set, but it hypothetically be promoted instead and lowered to:
+# fcvt s2, h0
+# fcvt s3, h1
+# fadd s2, s2, s3
+# fcvt h2, s2
+# which would depend on FZ instead and therefore not flush.
+#
+# However, making const-folding match the behaviour of the simple scalar version of these
+# instructions makes it more likely to be match the results from the hardware-executed
+# instruction than if we always ignore `denormal_fpenv` during const-folding.
+
+--- |
+ define void @fbinops() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @fma() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @ftz_boundary() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @fcmp() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @dynamic() denormal_fpenv(float: dynamic) { ret void }
+
+ define void @ceil() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @ceil_dynamic() denormal_fpenv(float: dynamic) { ret void }
+
+ define void @minmax() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @minmax_dynamic() denormal_fpenv(float: dynamic) { ret void }
+
+ define void @f16_ftz() denormal_fpenv(preservesign, float: ieee) { ret void }
+
+ define void @f16_dynamic() denormal_fpenv(dynamic, float: ieee) { ret void }
+
+ define void @f16_ieee() denormal_fpenv(float: preservesign) { ret void }
+
+ define void @f64_ftz() denormal_fpenv(preservesign) { ret void }
+
+ define void @f64_dynamic() denormal_fpenv(dynamic) { ret void }
+
+ define void @f64_ieee() denormal_fpenv(ieee) { ret void }
+
+...
+
+---
+name: fbinops
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; TODO: Flush the subnormal inputs and fold all four results to +0.
+ ; CHECK-LABEL: name: fbinops
+ ; CHECK: %add:_(f32) = G_FCONSTANT float 2.802600e-45
+ ; CHECK-NEXT: $s0 = COPY %add(f32)
+ ; CHECK-NEXT: $s1 = COPY %add(f32)
+ ; CHECK-NEXT: $s2 = COPY %add(f32)
+ ; CHECK-NEXT: $s3 = COPY %add(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %negdenorm:_(f32) = G_FCONSTANT float f0x80000001
+ %zero:_(f32) = G_FCONSTANT float 0.000000e+00
+ %two:_(f32) = G_FCONSTANT float 2.000000e+00
+ %half:_(f32) = G_FCONSTANT float 5.000000e-01
+ %add:_(f32) = G_FADD %denorm, %denorm
+ %sub:_(f32) = G_FSUB %denorm, %negdenorm
+ %mul:_(f32) = G_FMUL %denorm, %two
+ %div:_(f32) = G_FDIV %denorm, %half
+ $s0 = COPY %add
+ $s1 = COPY %sub
+ $s2 = COPY %mul
+ $s3 = COPY %div
+...
+
+---
+name: fma
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; TODO: Flush the subnormal input and fold G_FMA to +0.
+ ; CHECK-LABEL: name: fma
+ ; CHECK: %denorm:_(f32) = G_FCONSTANT float 1.401300e-45
+ ; CHECK-NEXT: $s0 = COPY %denorm(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %zero:_(f32) = G_FCONSTANT float 0.000000e+00
+ %one:_(f32) = G_FCONSTANT float 1.000000e+00
+ %result:_(f32) = G_FMA %denorm, %one, %zero
+ $s0 = COPY %result
+...
+
+---
+name: ftz_boundary
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; TODO: Leave G_FMUL and G_FMA unfolded rather than folding to min-normal.
+ ; The exact results are tiny and FZ can produce +0 instead of min-normal.
+ ; The exact behaviour depends on the AH flag, so we should not attempt
+ ; to model it during const-folding.
+ ; CHECK-LABEL: name: ftz_boundary
+ ; CHECK: %mul:_(f32) = G_FCONSTANT float f0x00800000
+ ; CHECK-NEXT: $s0 = COPY %mul(f32)
+ ; CHECK-NEXT: $s1 = COPY %mul(f32)
+ %a:_(f32) = G_FCONSTANT float f0x91FFFFFF
+ %b:_(f32) = G_FCONSTANT float f0xAE000000
+ %zero:_(f32) = G_FCONSTANT float 0.000000e+00
+ %mul:_(f32) = G_FMUL %a, %b
+ %fma:_(f32) = G_FMA %a, %b, %zero
+ $s0 = COPY %mul
+ $s1 = COPY %fma
+...
+
+---
+name: fcmp
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; TODO: Flush the subnormal input so G_FCMP is true and G_ZEXT yields 1.
+ ; CHECK-LABEL: name: fcmp
+ ; CHECK: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $w0 = COPY %extended(i32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %zero:_(f32) = G_FCONSTANT float 0.000000e+00
+ %result:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %result
+ $w0 = COPY %extended
+...
+
+---
+name: dynamic
+# denormal_fpenv(float: dynamic)
+body: |
+ bb.0:
+ ; TODO: Leave G_FADD unfolded: its result depends on dynamic flushing.
+ ; CHECK-LABEL: name: dynamic
+ ; CHECK: %result:_(f32) = G_FCONSTANT float 2.802600e-45
+ ; CHECK-NEXT: $s0 = COPY %result(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %result:_(f32) = G_FADD %denorm, %denorm
+ $s0 = COPY %result
+...
+
+---
+name: ceil
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; TODO: Flush the subnormal input and fold G_FCEIL to +0 instead of 1.
+ ; CHECK-LABEL: name: ceil
+ ; CHECK: %result:_(f32) = G_FCONSTANT float 1.000000e+00
+ ; CHECK-NEXT: $s0 = COPY %result(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %result:_(f32) = G_FCEIL %denorm
+ $s0 = COPY %result
+...
+
+---
+name: ceil_dynamic
+# denormal_fpenv(float: dynamic)
+body: |
+ bb.0:
+ ; TODO: Leave G_FCEIL unfolded: IEEE gives 1, but input FTZ gives +0.
+ ; CHECK-LABEL: name: ceil_dynamic
+ ; CHECK: %result:_(f32) = G_FCONSTANT float 1.000000e+00
+ ; CHECK-NEXT: $s0 = COPY %result(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %result:_(f32) = G_FCEIL %denorm
+ $s0 = COPY %result
+...
+
+---
+name: minmax
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; TODO: Flush the subnormal inputs and fold all six min/max results to +0.
+ ; CHECK-LABEL: name: minmax
+ ; CHECK: %denorm:_(f32) = G_FCONSTANT float 1.401300e-45
+ ; CHECK-NEXT: $s0 = COPY %denorm(f32)
+ ; CHECK-NEXT: $s1 = COPY %denorm(f32)
+ ; CHECK-NEXT: $s2 = COPY %denorm(f32)
+ ; CHECK-NEXT: $s3 = COPY %denorm(f32)
+ ; CHECK-NEXT: $s4 = COPY %denorm(f32)
+ ; CHECK-NEXT: $s5 = COPY %denorm(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %zero:_(f32) = G_FCONSTANT float 0.000000e+00
+ %one:_(f32) = G_FCONSTANT float 1.000000e+00
+ %minnum:_(f32) = G_FMINNUM %denorm, %one
+ %maxnum:_(f32) = G_FMAXNUM %denorm, %zero
+ %minimum:_(f32) = G_FMINIMUM %denorm, %one
+ %maximum:_(f32) = G_FMAXIMUM %denorm, %zero
+ %minimumnum:_(f32) = G_FMINIMUMNUM %denorm, %one
+ %maximumnum:_(f32) = G_FMAXIMUMNUM %denorm, %zero
+ $s0 = COPY %minnum
+ $s1 = COPY %maxnum
+ $s2 = COPY %minimum
+ $s3 = COPY %maximum
+ $s4 = COPY %minimumnum
+ $s5 = COPY %maximumnum
+...
+
+---
+name: minmax_dynamic
+# denormal_fpenv(float: dynamic)
+body: |
+ bb.0:
+ ; TODO: Leave G_FMINNUM unfolded: its result depends on dynamic flushing.
+ ; CHECK-LABEL: name: minmax_dynamic
+ ; CHECK: %denorm:_(f32) = G_FCONSTANT float 1.401300e-45
+ ; CHECK-NEXT: $s0 = COPY %denorm(f32)
+ %denorm:_(f32) = G_FCONSTANT float f0x00000001
+ %one:_(f32) = G_FCONSTANT float 1.000000e+00
+ %result:_(f32) = G_FMINNUM %denorm, %one
+ $s0 = COPY %result
+...
+
+---
+name: f16_ftz
+# denormal_fpenv(preservesign, float: ieee)
+body: |
+ bb.0:
+ ; TODO: Fold the arithmetic, FMA, ceil, and min/max results to +0, and
+ ; G_FCMP to true (G_ZEXT to 1). Flush %output_mul to +0 too: its inputs
+ ; are normal, but its result is subnormal. The float: ieee mode only
+ ; applies to f32 and must not disable flushing for f16.
+ ; CHECK-LABEL: name: f16_ftz
+ ; CHECK: %denorm:_(f16) = G_FCONSTANT half 5.960460e-08
+ ; CHECK-NEXT: %one:_(f16) = G_FCONSTANT half 1.000000e+00
+ ; CHECK-NEXT: %add:_(f16) = G_FCONSTANT half 1.192090e-07
+ ; CHECK-NEXT: %output_mul:_(f16) = G_FCONSTANT half 3.051760e-05
+ ; CHECK-NEXT: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $h0 = COPY %add(f16)
+ ; CHECK-NEXT: $h1 = COPY %add(f16)
+ ; CHECK-NEXT: $h2 = COPY %add(f16)
+ ; CHECK-NEXT: $h3 = COPY %add(f16)
+ ; CHECK-NEXT: $h5 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h6 = COPY %one(f16)
+ ; CHECK-NEXT: $h7 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h8 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h9 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h10 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h11 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h12 = COPY %denorm(f16)
+ ; CHECK-NEXT: $w13 = COPY %extended(i32)
+ ; CHECK-NEXT: $h14 = COPY %output_mul(f16)
+ %denorm:_(f16) = G_FCONSTANT half f0x0001
+ %zero:_(f16) = G_FCONSTANT half 0.000000e+00
+ %negdenorm:_(f16) = G_FCONSTANT half f0x8001
+ %one:_(f16) = G_FCONSTANT half 1.000000e+00
+ %two:_(f16) = G_FCONSTANT half 2.000000e+00
+ %half:_(f16) = G_FCONSTANT half 5.000000e-01
+ %add:_(f16) = G_FADD %denorm, %denorm
+ %sub:_(f16) = G_FSUB %denorm, %negdenorm
+ %mul:_(f16) = G_FMUL %denorm, %two
+ %div:_(f16) = G_FDIV %denorm, %half
+ %fma:_(f16) = G_FMA %denorm, %one, %zero
+ %ceil:_(f16) = G_FCEIL %denorm
+ %minnum:_(f16) = G_FMINNUM %denorm, %one
+ %maxnum:_(f16) = G_FMAXNUM %denorm, %zero
+ %minimum:_(f16) = G_FMINIMUM %denorm, %one
+ %maximum:_(f16) = G_FMAXIMUM %denorm, %zero
+ %minimumnum:_(f16) = G_FMINIMUMNUM %denorm, %one
+ %maximumnum:_(f16) = G_FMAXIMUMNUM %denorm, %zero
+ %minnormal:_(f16) = G_FCONSTANT half f0x0400
+ %output_mul:_(f16) = G_FMUL %minnormal, %half
+ %cmp:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %cmp
+ $h0 = COPY %add
+ $h1 = COPY %sub
+ $h2 = COPY %mul
+ $h3 = COPY %div
+ $h5 = COPY %fma
+ $h6 = COPY %ceil
+ $h7 = COPY %minnum
+ $h8 = COPY %maxnum
+ $h9 = COPY %minimum
+ $h10 = COPY %maximum
+ $h11 = COPY %minimumnum
+ $h12 = COPY %maximumnum
+ $w13 = COPY %extended
+ $h14 = COPY %output_mul
+...
+
+---
+name: f16_dynamic
+# denormal_fpenv(dynamic, float: ieee)
+body: |
+ bb.0:
+ ; TODO: Leave G_FADD, G_FMA, G_FCEIL, G_FMINNUM, G_FMUL, and G_FCMP
+ ; unfolded: their results depend on dynamic input/output flushing for f16,
+ ; independently of the float: ieee mode.
+ ; CHECK-LABEL: name: f16_dynamic
+ ; CHECK: %denorm:_(f16) = G_FCONSTANT half 5.960460e-08
+ ; CHECK-NEXT: %one:_(f16) = G_FCONSTANT half 1.000000e+00
+ ; CHECK-NEXT: %add:_(f16) = G_FCONSTANT half 1.192090e-07
+ ; CHECK-NEXT: %output_mul:_(f16) = G_FCONSTANT half 3.051760e-05
+ ; CHECK-NEXT: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $h0 = COPY %add(f16)
+ ; CHECK-NEXT: $h1 = COPY %denorm(f16)
+ ; CHECK-NEXT: $h2 = COPY %one(f16)
+ ; CHECK-NEXT: $h3 = COPY %denorm(f16)
+ ; CHECK-NEXT: $w4 = COPY %extended(i32)
+ ; CHECK-NEXT: $h5 = COPY %output_mul(f16)
+ %denorm:_(f16) = G_FCONSTANT half f0x0001
+ %zero:_(f16) = G_FCONSTANT half 0.000000e+00
+ %one:_(f16) = G_FCONSTANT half 1.000000e+00
+ %add:_(f16) = G_FADD %denorm, %denorm
+ %fma:_(f16) = G_FMA %denorm, %one, %zero
+ %ceil:_(f16) = G_FCEIL %denorm
+ %minnum:_(f16) = G_FMINNUM %denorm, %one
+ %minnormal:_(f16) = G_FCONSTANT half f0x0400
+ %half:_(f16) = G_FCONSTANT half 5.000000e-01
+ %output_mul:_(f16) = G_FMUL %minnormal, %half
+ %cmp:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %cmp
+ $h0 = COPY %add
+ $h1 = COPY %fma
+ $h2 = COPY %ceil
+ $h3 = COPY %minnum
+ $w4 = COPY %extended
+ $h5 = COPY %output_mul
+...
+
+---
+name: f16_ieee
+# denormal_fpenv(float: preservesign)
+body: |
+ bb.0:
+ ; f16 defaults to IEEE; the f32 FTZ setting must not affect these folds.
+ ; G_FADD produces f0x0002, G_FCEIL produces 1, and G_FCMP is false
+ ; CHECK-LABEL: name: f16_ieee
+ ; CHECK: %add:_(f16) = G_FCONSTANT half 1.192090e-07
+ ; CHECK-NEXT: %ceil:_(f16) = G_FCONSTANT half 1.000000e+00
+ ; CHECK-NEXT: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $h0 = COPY %add(f16)
+ ; CHECK-NEXT: $h1 = COPY %ceil(f16)
+ ; CHECK-NEXT: $w2 = COPY %extended(i32)
+ %denorm:_(f16) = G_FCONSTANT half f0x0001
+ %zero:_(f16) = G_FCONSTANT half 0.000000e+00
+ %add:_(f16) = G_FADD %denorm, %denorm
+ %ceil:_(f16) = G_FCEIL %denorm
+ %cmp:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %cmp
+ $h0 = COPY %add
+ $h1 = COPY %ceil
+ $w2 = COPY %extended
+...
+
+---
+name: f64_ftz
+# denormal_fpenv(preservesign)
+body: |
+ bb.0:
+ ; TODO: Fold the arithmetic, FMA, ceil, and min/max results to +0, and
+ ; G_FCMP to true (G_ZEXT to 1). Flush %output_mul to +0 too: its inputs
+ ; are normal, but its result is subnormal. FPCR.FZ enables flushing
+ ; for both f32 and f64.
+ ; CHECK-LABEL: name: f64_ftz
+ ; CHECK: %denorm:_(f64) = G_FCONSTANT double 4.940660e-324
+ ; CHECK-NEXT: %one:_(f64) = G_FCONSTANT double 1.000000e+00
+ ; CHECK-NEXT: %add:_(f64) = G_FCONSTANT double 9.881310e-324
+ ; CHECK-NEXT: %output_mul:_(f64) = G_FCONSTANT double f0x0008000000000000
+ ; CHECK-NEXT: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $d0 = COPY %add(f64)
+ ; CHECK-NEXT: $d1 = COPY %add(f64)
+ ; CHECK-NEXT: $d2 = COPY %add(f64)
+ ; CHECK-NEXT: $d3 = COPY %add(f64)
+ ; CHECK-NEXT: $d5 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d6 = COPY %one(f64)
+ ; CHECK-NEXT: $d7 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d8 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d9 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d10 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d11 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d12 = COPY %denorm(f64)
+ ; CHECK-NEXT: $w26 = COPY %extended(i32)
+ ; CHECK-NEXT: $d14 = COPY %output_mul(f64)
+ %denorm:_(f64) = G_FCONSTANT double f0x0000000000000001
+ %zero:_(f64) = G_FCONSTANT double 0.000000e+00
+ %negdenorm:_(f64) = G_FCONSTANT double f0x8000000000000001
+ %one:_(f64) = G_FCONSTANT double 1.000000e+00
+ %two:_(f64) = G_FCONSTANT double 2.000000e+00
+ %half:_(f64) = G_FCONSTANT double 5.000000e-01
+ %add:_(f64) = G_FADD %denorm, %denorm
+ %sub:_(f64) = G_FSUB %denorm, %negdenorm
+ %mul:_(f64) = G_FMUL %denorm, %two
+ %div:_(f64) = G_FDIV %denorm, %half
+ %fma:_(f64) = G_FMA %denorm, %one, %zero
+ %ceil:_(f64) = G_FCEIL %denorm
+ %minnum:_(f64) = G_FMINNUM %denorm, %one
+ %maxnum:_(f64) = G_FMAXNUM %denorm, %zero
+ %minimum:_(f64) = G_FMINIMUM %denorm, %one
+ %maximum:_(f64) = G_FMAXIMUM %denorm, %zero
+ %minimumnum:_(f64) = G_FMINIMUMNUM %denorm, %one
+ %maximumnum:_(f64) = G_FMAXIMUMNUM %denorm, %zero
+ %minnormal:_(f64) = G_FCONSTANT double f0x0010000000000000
+ %output_mul:_(f64) = G_FMUL %minnormal, %half
+ %cmp:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %cmp
+ $d0 = COPY %add
+ $d1 = COPY %sub
+ $d2 = COPY %mul
+ $d3 = COPY %div
+ $d5 = COPY %fma
+ $d6 = COPY %ceil
+ $d7 = COPY %minnum
+ $d8 = COPY %maxnum
+ $d9 = COPY %minimum
+ $d10 = COPY %maximum
+ $d11 = COPY %minimumnum
+ $d12 = COPY %maximumnum
+ $w26 = COPY %extended
+ $d14 = COPY %output_mul
+...
+
+---
+name: f64_dynamic
+# denormal_fpenv(dynamic)
+body: |
+ bb.0:
+ ; TODO: Leave G_FADD, G_FMA, G_FCEIL, G_FMINNUM, G_FMUL, and G_FCMP
+ ; unfolded: their results depend on dynamic input/output flushing.
+ ; The same FPCR.FZ setting controls f32 and f64.
+ ; CHECK-LABEL: name: f64_dynamic
+ ; CHECK: %denorm:_(f64) = G_FCONSTANT double 4.940660e-324
+ ; CHECK-NEXT: %one:_(f64) = G_FCONSTANT double 1.000000e+00
+ ; CHECK-NEXT: %add:_(f64) = G_FCONSTANT double 9.881310e-324
+ ; CHECK-NEXT: %output_mul:_(f64) = G_FCONSTANT double f0x0008000000000000
+ ; CHECK-NEXT: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $d0 = COPY %add(f64)
+ ; CHECK-NEXT: $d1 = COPY %denorm(f64)
+ ; CHECK-NEXT: $d2 = COPY %one(f64)
+ ; CHECK-NEXT: $d3 = COPY %denorm(f64)
+ ; CHECK-NEXT: $w8 = COPY %extended(i32)
+ ; CHECK-NEXT: $d5 = COPY %output_mul(f64)
+ %denorm:_(f64) = G_FCONSTANT double f0x0000000000000001
+ %zero:_(f64) = G_FCONSTANT double 0.000000e+00
+ %one:_(f64) = G_FCONSTANT double 1.000000e+00
+ %add:_(f64) = G_FADD %denorm, %denorm
+ %fma:_(f64) = G_FMA %denorm, %one, %zero
+ %ceil:_(f64) = G_FCEIL %denorm
+ %minnum:_(f64) = G_FMINNUM %denorm, %one
+ %minnormal:_(f64) = G_FCONSTANT double f0x0010000000000000
+ %half:_(f64) = G_FCONSTANT double 5.000000e-01
+ %output_mul:_(f64) = G_FMUL %minnormal, %half
+ %cmp:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %cmp
+ $d0 = COPY %add
+ $d1 = COPY %fma
+ $d2 = COPY %ceil
+ $d3 = COPY %minnum
+ $w8 = COPY %extended
+ $d5 = COPY %output_mul
+...
+
+---
+name: f64_ieee
+# denormal_fpenv(ieee)
+body: |
+ bb.0:
+ ; With FPCR.FZ disabled, both f32 and f64 use IEEE denormal handling.
+ ; G_FADD produces f0x0000000000000002, G_FCEIL produces 1, and G_FCMP is false
+ ; CHECK-LABEL: name: f64_ieee
+ ; CHECK: %add:_(f64) = G_FCONSTANT double 9.881310e-324
+ ; CHECK-NEXT: %ceil:_(f64) = G_FCONSTANT double 1.000000e+00
+ ; CHECK-NEXT: %extended:_(i32) = G_CONSTANT i32 0
+ ; CHECK-NEXT: $d0 = COPY %add(f64)
+ ; CHECK-NEXT: $d1 = COPY %ceil(f64)
+ ; CHECK-NEXT: $w4 = COPY %extended(i32)
+ %denorm:_(f64) = G_FCONSTANT double f0x0000000000000001
+ %zero:_(f64) = G_FCONSTANT double 0.000000e+00
+ %add:_(f64) = G_FADD %denorm, %denorm
+ %ceil:_(f64) = G_FCEIL %denorm
+ %cmp:_(i1) = G_FCMP floatpred(oeq), %denorm, %zero
+ %extended:_(i32) = G_ZEXT %cmp
+ $d0 = COPY %add
+ $d1 = COPY %ceil
+ $w4 = COPY %extended
+...
>From 13378194f30023cf1b8980cd56ecb4c04e0d509a Mon Sep 17 00:00:00 2001
From: Lewis Crawford <lcrawford at nvidia.com>
Date: Sat, 21 Nov 2026 12:13:25 +0000
Subject: [PATCH 19/35] [ConstFold][NFC] Add IR denorm misfolding baseline
(#236190)
This patch adds baseline tests with FIXME comments to highlight results
that are missing denormal_fpenv-aware flushing during const-folding with,
which will be addressed in a follow-up patch.
Currently ConstantFolding.cpp is the only IR code-path that tries to handle
denormal_fpenv while const-folding, and even there it misses several
cases like FMA and fmax. This results in inconsistently folded results in
several cases with a non-IEEE denormal_fpenv.
The cases ignoring denormal_fpenv that are covered in these new baseline tests are:
- InstSimplify arithmetic: FMA, rounding, comparisons, and min/max.
- Constant folding during InstCombine reassociation and factorization.
- BranchProbabilityInfo folding of loop-carried value chains.
- NVVM multiply, divide, and FMA at the subnormal rounding boundary.
---
.../denormal-fcmp-folding.ll | 39 ++++++
.../denormal-reassociation-folding.ll | 125 ++++++++++++++++++
.../InstSimplify/denormal-folding.ll | 79 +++++++++++
.../InstSimplify/nvvm-ftz-boundary-folding.ll | 51 +++++++
4 files changed, 294 insertions(+)
create mode 100644 llvm/test/Analysis/BranchProbabilityInfo/denormal-fcmp-folding.ll
create mode 100644 llvm/test/Transforms/InstCombine/denormal-reassociation-folding.ll
create mode 100644 llvm/test/Transforms/InstSimplify/denormal-folding.ll
create mode 100644 llvm/test/Transforms/InstSimplify/nvvm-ftz-boundary-folding.ll
diff --git a/llvm/test/Analysis/BranchProbabilityInfo/denormal-fcmp-folding.ll b/llvm/test/Analysis/BranchProbabilityInfo/denormal-fcmp-folding.ll
new file mode 100644
index 00000000000000..4b3e32f7683780
--- /dev/null
+++ b/llvm/test/Analysis/BranchProbabilityInfo/denormal-fcmp-folding.ll
@@ -0,0 +1,39 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -passes='print<branch-prob>' -disable-output < %s 2>&1 | FileCheck %s
+
+; BPI folds the operation chain from the loop-carried PHI to the branch
+; condition while identifying an unlikely loop-reset edge.
+; FIXME: Honor the function's denormal environment when folding this chain.
+; The reset supplies a subnormal, which should flush to zero. The `one`
+; comparison is then false on the next iteration, making %reset unlikely
+; (33.33%) instead of the current 62.50% estimate.
+
+define void @fadd_chain(float %unknown) denormal_fpenv(float: preservesign) {
+; CHECK-LABEL: 'fadd_chain'
+; CHECK-NEXT: ---- Branch Probabilities ----
+; CHECK-NEXT: edge %entry -> %header probability is 0x80000000 / 0x80000000 = 100.00% [HOT edge]
+; CHECK-NEXT: edge %header -> %body probability is 0x80000000 / 0x80000000 = 100.00% [HOT edge]
+; CHECK-NEXT: edge %body -> %reset probability is 0x50000000 / 0x80000000 = 62.50%
+; CHECK-NEXT: edge %body -> %latch probability is 0x30000000 / 0x80000000 = 37.50%
+; CHECK-NEXT: edge %reset -> %latch probability is 0x80000000 / 0x80000000 = 100.00% [HOT edge]
+; CHECK-NEXT: edge %latch -> %header probability is 0x80000000 / 0x80000000 = 100.00% [HOT edge]
+;
+entry:
+ br label %header
+
+header:
+ %value = phi float [ 0.000000e+00, %entry ], [ %next, %latch ]
+ br label %body
+
+body:
+ %sum = fadd float %value, 0.000000e+00
+ %cmp = fcmp one float %sum, 0.000000e+00
+ br i1 %cmp, label %reset, label %latch
+
+reset:
+ br label %latch
+
+latch:
+ %next = phi float [ f0x00000001, %reset ], [ %unknown, %body ]
+ br label %header
+}
diff --git a/llvm/test/Transforms/InstCombine/denormal-reassociation-folding.ll b/llvm/test/Transforms/InstCombine/denormal-reassociation-folding.ll
new file mode 100644
index 00000000000000..8e863ee3e53b54
--- /dev/null
+++ b/llvm/test/Transforms/InstCombine/denormal-reassociation-folding.ll
@@ -0,0 +1,125 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -passes=instcombine -S < %s | FileCheck %s
+
+; Record current constant folding during reassociation and factorization.
+; FIXME: Honor the function's denormal environment when folding constants
+; inside these transformations.
+
+; FIXME: Flush both constants to +0 so the result reduces to fadd fast %x, %y,
+; instead of retaining an addition of f0x00000002.
+define float @reassociate_fadd(float %x, float %y) denormal_fpenv(float: preservesign) {
+; CHECK-LABEL: define float @reassociate_fadd(
+; CHECK-SAME: float [[X:%.*]], float [[Y:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: [[RESULT:%.*]] = fadd fast float [[X]], [[Y]]
+; CHECK-NEXT: [[RESULT1:%.*]] = fadd fast float [[RESULT]], 2.802600e-45
+; CHECK-NEXT: ret float [[RESULT1]]
+;
+ %x.add = fadd fast float %x, f0x00000001
+ %y.add = fadd fast float %y, f0x00000001
+ %result = fadd fast float %x.add, %y.add
+ ret float %result
+}
+
+; FIXME: Flush the combined constants' subnormal result to +0 so the result
+; reduces to fadd fast %x, %y, instead of retaining an addition of f0x00000001.
+define float @reassociate_output_ftz(float %x, float %y) denormal_fpenv(float: preservesign|ieee) {
+; CHECK-LABEL: define float @reassociate_output_ftz(
+; CHECK-SAME: float [[X:%.*]], float [[Y:%.*]]) #[[ATTR1:[0-9]+]] {
+; CHECK-NEXT: [[Y_ADD:%.*]] = fadd fast float [[X]], [[Y]]
+; CHECK-NEXT: [[RESULT:%.*]] = fadd fast float [[Y_ADD]], 1.401300e-45
+; CHECK-NEXT: ret float [[RESULT]]
+;
+ %x.add = fadd fast float %x, f0x00800000
+ %y.add = fadd fast float %y, f0x807FFFFF
+ %result = fadd fast float %x.add, %y.add
+ ret float %result
+}
+
+; FIXME: Preserve the original additions instead of combining the constants:
+; their sum depends on the dynamic denormal environment.
+define float @reassociate_dynamic(float %x, float %y) denormal_fpenv(float: dynamic) {
+; CHECK-LABEL: define float @reassociate_dynamic(
+; CHECK-SAME: float [[X:%.*]], float [[Y:%.*]]) #[[ATTR2:[0-9]+]] {
+; CHECK-NEXT: [[Y_ADD:%.*]] = fadd fast float [[X]], [[Y]]
+; CHECK-NEXT: [[RESULT:%.*]] = fadd fast float [[Y_ADD]], 2.802600e-45
+; CHECK-NEXT: ret float [[RESULT]]
+;
+ %x.add = fadd fast float %x, f0x00000001
+ %y.add = fadd fast float %y, f0x00000001
+ %result = fadd fast float %x.add, %y.add
+ ret float %result
+}
+
+; Factorization creates (C1 + C2) / X. With input FTZ, the subnormal C2 is zero.
+; FIXME: Fold the numerator to C1 (f0x80800000), enabling factorization to a
+; single fdiv fast float f0x80800000, %x.
+define float @factor_fdiv_input_ftz(float %x) denormal_fpenv(float: ieee|preservesign) {
+; CHECK-LABEL: define float @factor_fdiv_input_ftz(
+; CHECK-SAME: float [[X:%.*]]) #[[ATTR3:[0-9]+]] {
+; CHECK-NEXT: [[R:%.*]] = fdiv fast float f0x80800000, [[X]]
+; CHECK-NEXT: [[RHS:%.*]] = fdiv fast float f0x00400000, [[X]]
+; CHECK-NEXT: [[RESULT:%.*]] = fadd fast float [[R]], [[RHS]]
+; CHECK-NEXT: ret float [[RESULT]]
+;
+ %lhs = fdiv fast float f0x80800000, %x
+ %rhs = fdiv fast float f0x00400000, %x
+ %result = fadd fast float %lhs, %rhs
+ ret float %result
+}
+
+; Factorization correctly retains the original operations here. Folding C1 + C2
+; depends on dynamic input flushing, so keep declining this transformation
+; once constant folding becomes denormal-aware.
+define float @factor_fdiv_dynamic(float %x) denormal_fpenv(float: dynamic) {
+; CHECK-LABEL: define float @factor_fdiv_dynamic(
+; CHECK-SAME: float [[X:%.*]]) #[[ATTR2]] {
+; CHECK-NEXT: [[LHS:%.*]] = fdiv fast float f0x80800000, [[X]]
+; CHECK-NEXT: [[RHS:%.*]] = fdiv fast float f0x00400000, [[X]]
+; CHECK-NEXT: [[RESULT:%.*]] = fadd fast float [[LHS]], [[RHS]]
+; CHECK-NEXT: ret float [[RESULT]]
+;
+ %lhs = fdiv fast float f0x80800000, %x
+ %rhs = fdiv fast float f0x00400000, %x
+ %result = fadd fast float %lhs, %rhs
+ ret float %result
+}
+
+; FIXME: Use +0 for the combined bound in all six min/max intrinsics instead
+; of retaining min-subnormal, since the folded constant inputs should flush.
+define void @minmax_reassociate(float %x, ptr %minnum_res, ptr %maxnum_res, ptr %minimum_res, ptr %maximum_res, ptr %minimumnum_res, ptr %maximumnum_res) denormal_fpenv(float: preservesign) {
+; CHECK-LABEL: define void @minmax_reassociate(
+; CHECK-SAME: float [[X:%.*]], ptr [[MINNUM_RES:%.*]], ptr [[MAXNUM_RES:%.*]], ptr [[MINIMUM_RES:%.*]], ptr [[MAXIMUM_RES:%.*]], ptr [[MINIMUMNUM_RES:%.*]], ptr [[MAXIMUMNUM_RES:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[MINNUM_1:%.*]] = call float @llvm.minnum.f32(float [[X]], float 1.401300e-45)
+; CHECK-NEXT: store float [[MINNUM_1]], ptr [[MINNUM_RES]], align 4
+; CHECK-NEXT: [[MAXNUM_1:%.*]] = call float @llvm.maxnum.f32(float [[X]], float 1.401300e-45)
+; CHECK-NEXT: store float [[MAXNUM_1]], ptr [[MAXNUM_RES]], align 4
+; CHECK-NEXT: [[MINIMUM_1:%.*]] = call float @llvm.minimum.f32(float [[X]], float 1.401300e-45)
+; CHECK-NEXT: store float [[MINIMUM_1]], ptr [[MINIMUM_RES]], align 4
+; CHECK-NEXT: [[MAXIMUM_1:%.*]] = call float @llvm.maximum.f32(float [[X]], float 1.401300e-45)
+; CHECK-NEXT: store float [[MAXIMUM_1]], ptr [[MAXIMUM_RES]], align 4
+; CHECK-NEXT: [[MINIMUMNUM_1:%.*]] = call float @llvm.minimumnum.f32(float [[X]], float 1.401300e-45)
+; CHECK-NEXT: store float [[MINIMUMNUM_1]], ptr [[MINIMUMNUM_RES]], align 4
+; CHECK-NEXT: [[MAXIMUMNUM_1:%.*]] = call float @llvm.maximumnum.f32(float [[X]], float 1.401300e-45)
+; CHECK-NEXT: store float [[MAXIMUMNUM_1]], ptr [[MAXIMUMNUM_RES]], align 4
+; CHECK-NEXT: ret void
+;
+ %minnum.0 = call float @llvm.minnum.f32(float %x, float f0x00000001)
+ %minnum.1 = call float @llvm.minnum.f32(float %minnum.0, float 1.000000e+00)
+ store float %minnum.1, ptr %minnum_res
+ %maxnum.0 = call float @llvm.maxnum.f32(float %x, float f0x00000001)
+ %maxnum.1 = call float @llvm.maxnum.f32(float %maxnum.0, float 0.000000e+00)
+ store float %maxnum.1, ptr %maxnum_res
+ %minimum.0 = call float @llvm.minimum.f32(float %x, float f0x00000001)
+ %minimum.1 = call float @llvm.minimum.f32(float %minimum.0, float 1.000000e+00)
+ store float %minimum.1, ptr %minimum_res
+ %maximum.0 = call float @llvm.maximum.f32(float %x, float f0x00000001)
+ %maximum.1 = call float @llvm.maximum.f32(float %maximum.0, float 0.000000e+00)
+ store float %maximum.1, ptr %maximum_res
+ %minimumnum.0 = call float @llvm.minimumnum.f32(float %x, float f0x00000001)
+ %minimumnum.1 = call float @llvm.minimumnum.f32(float %minimumnum.0, float 1.000000e+00)
+ store float %minimumnum.1, ptr %minimumnum_res
+ %maximumnum.0 = call float @llvm.maximumnum.f32(float %x, float f0x00000001)
+ %maximumnum.1 = call float @llvm.maximumnum.f32(float %maximumnum.0, float 0.000000e+00)
+ store float %maximumnum.1, ptr %maximumnum_res
+ ret void
+}
diff --git a/llvm/test/Transforms/InstSimplify/denormal-folding.ll b/llvm/test/Transforms/InstSimplify/denormal-folding.ll
new file mode 100644
index 00000000000000..9cc04780c5eead
--- /dev/null
+++ b/llvm/test/Transforms/InstSimplify/denormal-folding.ll
@@ -0,0 +1,79 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -passes=instsimplify -S < %s | FileCheck %s
+
+; Record current IR constant folding under different denormal environments.
+; FIXME: Honor the function's denormal environment in every constant fold.
+
+; Scalar binary arithmetic, vector addition, and fcmp already honor input FTZ.
+; FIXME: fmuladd, fma, ceil, and all six min/max variants should also store +0.
+; In particular, ceil(min-subnormal) should use a flushed input, producing +0
+; instead of 1.
+define void @preservesign(ptr %fadd_res, ptr %fsub_res, ptr %fmul_res, ptr %fdiv_res, ptr %fmuladd_res, ptr %fma_res, ptr %ceil_res, ptr %fcmp_res, ptr %vector_add_res, ptr %minnum_res, ptr %maxnum_res, ptr %minimum_res, ptr %maximum_res, ptr %minimumnum_res, ptr %maximumnum_res) denormal_fpenv(float: preservesign) {
+; CHECK-LABEL: define void @preservesign(
+; CHECK-SAME: ptr [[FADD_RES:%.*]], ptr [[FSUB_RES:%.*]], ptr [[FMUL_RES:%.*]], ptr [[FDIV_RES:%.*]], ptr [[FMULADD_RES:%.*]], ptr [[FMA_RES:%.*]], ptr [[CEIL_RES:%.*]], ptr [[FCMP_RES:%.*]], ptr [[VECTOR_ADD_RES:%.*]], ptr [[MINNUM_RES:%.*]], ptr [[MAXNUM_RES:%.*]], ptr [[MINIMUM_RES:%.*]], ptr [[MAXIMUM_RES:%.*]], ptr [[MINIMUMNUM_RES:%.*]], ptr [[MAXIMUMNUM_RES:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: store float 0.000000e+00, ptr [[FADD_RES]], align 4
+; CHECK-NEXT: store float 0.000000e+00, ptr [[FSUB_RES]], align 4
+; CHECK-NEXT: store float 0.000000e+00, ptr [[FMUL_RES]], align 4
+; CHECK-NEXT: store float 0.000000e+00, ptr [[FDIV_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[FMULADD_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[FMA_RES]], align 4
+; CHECK-NEXT: store float 1.000000e+00, ptr [[CEIL_RES]], align 4
+; CHECK-NEXT: store i1 true, ptr [[FCMP_RES]], align 1
+; CHECK-NEXT: store <2 x float> zeroinitializer, ptr [[VECTOR_ADD_RES]], align 8
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MINNUM_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MAXNUM_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MINIMUM_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MAXIMUM_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MINIMUMNUM_RES]], align 4
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MAXIMUMNUM_RES]], align 4
+; CHECK-NEXT: ret void
+;
+ %fadd = fadd float f0x00000001, f0x00000001
+ store float %fadd, ptr %fadd_res
+ %fsub = fsub float f0x00000001, f0x80000001
+ store float %fsub, ptr %fsub_res
+ %fmul = fmul float f0x00000001, 2.0
+ store float %fmul, ptr %fmul_res
+ %fdiv = fdiv float f0x00000001, 0.5
+ store float %fdiv, ptr %fdiv_res
+ %fmuladd = call float @llvm.fmuladd.f32(float f0x00000001, float 1.0, float 0.0)
+ store float %fmuladd, ptr %fmuladd_res
+ %fma = call float @llvm.fma.f32(float f0x00000001, float 1.0, float 0.0)
+ store float %fma, ptr %fma_res
+ %ceil = call float @llvm.ceil.f32(float f0x00000001)
+ store float %ceil, ptr %ceil_res
+ %fcmp = fcmp oeq float f0x00000001, 0.0
+ store i1 %fcmp, ptr %fcmp_res
+ %vector_add = fadd <2 x float> <float f0x00000001, float 0.0>, <float 0.0, float f0x00000001>
+ store <2 x float> %vector_add, ptr %vector_add_res
+ %minnum = call float @llvm.minnum.f32(float f0x00000001, float 1.0)
+ store float %minnum, ptr %minnum_res
+ %maxnum = call float @llvm.maxnum.f32(float f0x00000001, float 0.0)
+ store float %maxnum, ptr %maxnum_res
+ %minimum = call float @llvm.minimum.f32(float f0x00000001, float 1.0)
+ store float %minimum, ptr %minimum_res
+ %maximum = call float @llvm.maximum.f32(float f0x00000001, float 0.0)
+ store float %maximum, ptr %maximum_res
+ %minimumnum = call float @llvm.minimumnum.f32(float f0x00000001, float 1.0)
+ store float %minimumnum, ptr %minimumnum_res
+ %maximumnum = call float @llvm.maximumnum.f32(float f0x00000001, float 0.0)
+ store float %maximumnum, ptr %maximumnum_res
+ ret void
+}
+
+; fmul correctly remains unfolded because its result depends on dynamic FTZ.
+; FIXME: Keep minnum unfolded too: its result can be min-subnormal or +0.
+define void @dynamic(ptr %minnum_res, ptr %fmul_res) denormal_fpenv(float: dynamic) {
+; CHECK-LABEL: define void @dynamic(
+; CHECK-SAME: ptr [[MINNUM_RES:%.*]], ptr [[FMUL_RES:%.*]]) #[[ATTR1:[0-9]+]] {
+; CHECK-NEXT: store float 1.401300e-45, ptr [[MINNUM_RES]], align 4
+; CHECK-NEXT: [[FMUL:%.*]] = fmul float 1.401300e-45, 2.000000e+00
+; CHECK-NEXT: store float [[FMUL]], ptr [[FMUL_RES]], align 4
+; CHECK-NEXT: ret void
+;
+ %minnum = call float @llvm.minnum.f32(float f0x00000001, float 1.0)
+ store float %minnum, ptr %minnum_res
+ %fmul = fmul float f0x00000001, 2.0
+ store float %fmul, ptr %fmul_res
+ ret void
+}
diff --git a/llvm/test/Transforms/InstSimplify/nvvm-ftz-boundary-folding.ll b/llvm/test/Transforms/InstSimplify/nvvm-ftz-boundary-folding.ll
new file mode 100644
index 00000000000000..ecf58e5fb19989
--- /dev/null
+++ b/llvm/test/Transforms/InstSimplify/nvvm-ftz-boundary-folding.ll
@@ -0,0 +1,51 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -passes=instsimplify -mtriple=nvptx64 -S < %s | FileCheck %s
+
+; Record current folding at the FTZ/gradual-underflow rounding boundary.
+; FIXME: Leave these calls unfolded. Their exact results are subnormal and
+; native FTZ produces signed zero before rounding, but the current folders
+; incorrectly round them up to positive/negative minimum-normal.
+
+define void @nvvm_ftz_boundaries(ptr %mul_rn_res, ptr %mul_rp_res, ptr %mul_negative_rn_res, ptr %mul_negative_rm_res, ptr %div_rn_res, ptr %div_rp_res, ptr %div_negative_rn_res, ptr %div_negative_rm_res, ptr %fma_rn_res, ptr %fma_rp_res, ptr %fma_negative_rn_res, ptr %fma_negative_rm_res) {
+; CHECK-LABEL: define void @nvvm_ftz_boundaries(
+; CHECK-SAME: ptr [[MUL_RN_RES:%.*]], ptr [[MUL_RP_RES:%.*]], ptr [[MUL_NEGATIVE_RN_RES:%.*]], ptr [[MUL_NEGATIVE_RM_RES:%.*]], ptr [[DIV_RN_RES:%.*]], ptr [[DIV_RP_RES:%.*]], ptr [[DIV_NEGATIVE_RN_RES:%.*]], ptr [[DIV_NEGATIVE_RM_RES:%.*]], ptr [[FMA_RN_RES:%.*]], ptr [[FMA_RP_RES:%.*]], ptr [[FMA_NEGATIVE_RN_RES:%.*]], ptr [[FMA_NEGATIVE_RM_RES:%.*]]) {
+; CHECK-NEXT: store float f0x00800000, ptr [[MUL_RN_RES]], align 4
+; CHECK-NEXT: store float f0x00800000, ptr [[MUL_RP_RES]], align 4
+; CHECK-NEXT: store float f0x80800000, ptr [[MUL_NEGATIVE_RN_RES]], align 4
+; CHECK-NEXT: store float f0x80800000, ptr [[MUL_NEGATIVE_RM_RES]], align 4
+; CHECK-NEXT: store float f0x00800000, ptr [[DIV_RN_RES]], align 4
+; CHECK-NEXT: store float f0x00800000, ptr [[DIV_RP_RES]], align 4
+; CHECK-NEXT: store float f0x80800000, ptr [[DIV_NEGATIVE_RN_RES]], align 4
+; CHECK-NEXT: store float f0x80800000, ptr [[DIV_NEGATIVE_RM_RES]], align 4
+; CHECK-NEXT: store float f0x00800000, ptr [[FMA_RN_RES]], align 4
+; CHECK-NEXT: store float f0x00800000, ptr [[FMA_RP_RES]], align 4
+; CHECK-NEXT: store float f0x80800000, ptr [[FMA_NEGATIVE_RN_RES]], align 4
+; CHECK-NEXT: store float f0x80800000, ptr [[FMA_NEGATIVE_RM_RES]], align 4
+; CHECK-NEXT: ret void
+;
+ %mul_rn = call float @llvm.nvvm.fmul.ftz.f32(float f0x00800000, float f0x3F7FFFFF, i32 1)
+ store float %mul_rn, ptr %mul_rn_res
+ %mul_rp = call float @llvm.nvvm.fmul.ftz.f32(float f0x00800000, float f0x3F7FFFFF, i32 2)
+ store float %mul_rp, ptr %mul_rp_res
+ %mul_negative_rn = call float @llvm.nvvm.fmul.ftz.f32(float f0x80800000, float f0x3F7FFFFF, i32 1)
+ store float %mul_negative_rn, ptr %mul_negative_rn_res
+ %mul_negative_rm = call float @llvm.nvvm.fmul.ftz.f32(float f0x80800000, float f0x3F7FFFFF, i32 3)
+ store float %mul_negative_rm, ptr %mul_negative_rm_res
+ %div_rn = call float @llvm.nvvm.div.rn.ftz.f(float f0x00FFFFFF, float 2.000000e+00)
+ store float %div_rn, ptr %div_rn_res
+ %div_rp = call float @llvm.nvvm.div.rp.ftz.f(float f0x00800000, float f0x3F800001)
+ store float %div_rp, ptr %div_rp_res
+ %div_negative_rn = call float @llvm.nvvm.div.rn.ftz.f(float f0x80FFFFFF, float 2.000000e+00)
+ store float %div_negative_rn, ptr %div_negative_rn_res
+ %div_negative_rm = call float @llvm.nvvm.div.rm.ftz.f(float f0x80800000, float f0x3F800001)
+ store float %div_negative_rm, ptr %div_negative_rm_res
+ %fma_rn = call float @llvm.nvvm.fma.rn.ftz.f(float f0x00800000, float f0x3F7FFFFF, float 0.000000e+00)
+ store float %fma_rn, ptr %fma_rn_res
+ %fma_rp = call float @llvm.nvvm.fma.rp.ftz.f(float f0x00800000, float f0x3F7FFFFF, float 0.000000e+00)
+ store float %fma_rp, ptr %fma_rp_res
+ %fma_negative_rn = call float @llvm.nvvm.fma.rn.ftz.f(float f0x80800000, float f0x3F7FFFFF, float 0.000000e+00)
+ store float %fma_negative_rn, ptr %fma_negative_rn_res
+ %fma_negative_rm = call float @llvm.nvvm.fma.rm.ftz.f(float f0x80800000, float f0x3F7FFFFF, float 0.000000e+00)
+ store float %fma_negative_rm, ptr %fma_negative_rm_res
+ ret void
+}
>From fac5fe597904f70d9fdcf872b2ff6a254a1a2e79 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Sat, 21 Nov 2026 12:47:07 +0000
Subject: [PATCH 20/35] [X86] lowerShuffleAsZeroOrAnyExtend - ensure we bitcast
VZEXT_MOVL source to the result type (#244646)
lowerShuffleAsZeroOrAnyExtend can be used with v4i32 and v4f32 types,
but we expect VZEXT_MOVL to be used with v4i32 - isel was still working,
but improvements to SDNode verification was causing an assert
Fixes #244568
---
llvm/lib/Target/X86/X86ISelLowering.cpp | 2 +-
.../CodeGen/X86/vector-shuffle-combining.ll | 31 +++++++++++++++++--
2 files changed, 30 insertions(+), 3 deletions(-)
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index b269f3a4a1ffe4..60450e5a63f212 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -13076,7 +13076,7 @@ static SDValue lowerShuffleAsZeroOrAnyExtend(
if (VT == MVT::v4i32 &&
(V1.getOpcode() == ISD::SCALAR_TO_VECTOR || isa<MemSDNode>(V1)) &&
Mask[0] == 0 && (NumElements - 1) == (int)Zeroable.countLeadingOnes())
- return DAG.getNode(X86ISD::VZEXT_MOVL, DL, VT, V1);
+ return DAG.getNode(X86ISD::VZEXT_MOVL, DL, VT, DAG.getBitcast(VT, V1));
// The widest scale possible for extending is to a 64-bit integer.
assert(Bits % 64 == 0 &&
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
index 4435515ae42efa..61e4df0d647be8 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -2901,6 +2901,33 @@ define <4 x i32> @PR41545(<4 x i32> %a0, <16 x i8> %a1) {
ret <4 x i32> %15
}
+define <8 x float> @PR244568(ptr %p) {
+; SSE2-LABEL: PR244568:
+; SSE2: # %bb.0:
+; SSE2-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSE2-NEXT: retq
+;
+; SSSE3-LABEL: PR244568:
+; SSSE3: # %bb.0:
+; SSSE3-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSSE3-NEXT: retq
+;
+; SSE41-LABEL: PR244568:
+; SSE41: # %bb.0:
+; SSE41-NEXT: xorps %xmm1, %xmm1
+; SSE41-NEXT: blendps {{.*#+}} xmm1 = mem[0,1],xmm1[2],mem[3]
+; SSE41-NEXT: retq
+;
+; AVX-LABEL: PR244568:
+; AVX: # %bb.0:
+; AVX-NEXT: vxorps %xmm0, %xmm0, %xmm0
+; AVX-NEXT: vblendps {{.*#+}} ymm0 = mem[0,1,2,3,4,5],ymm0[6,7]
+; AVX-NEXT: retq
+ %v = load <8 x float>, ptr %p
+ %r = shufflevector <8 x float> %v, <8 x float> <float poison, float poison, float poison, float poison, float poison, float poison, float poison, float 0.0>, <8 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 4, i32 poison, i32 15, i32 poison>
+ ret <8 x float> %r
+}
+
define <8 x i16> @shuffle_extract_insert(<8 x i16> %a) {
; SSE-LABEL: shuffle_extract_insert:
; SSE: # %bb.0:
@@ -3779,9 +3806,9 @@ define void @autogen_SD25931() {
; CHECK-LABEL: autogen_SD25931:
; CHECK: # %bb.0: # %BB
; CHECK-NEXT: .p2align 4
-; CHECK-NEXT: .LBB145_1: # %CF242
+; CHECK-NEXT: .LBB146_1: # %CF242
; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: jmp .LBB145_1
+; CHECK-NEXT: jmp .LBB146_1
BB:
%Cmp16 = icmp uge <2 x i1> zeroinitializer, zeroinitializer
%Shuff19 = shufflevector <2 x i1> zeroinitializer, <2 x i1> %Cmp16, <2 x i32> <i32 3, i32 1>
>From 87764a902a9bdc02c52818ec4f226b962843f9c1 Mon Sep 17 00:00:00 2001
From: Phoebe Wang <phoebe.wang at intel.com>
Date: Sat, 21 Nov 2026 20:47:41 +0800
Subject: [PATCH 21/35] [X86][AVX10.2] Fix mask type and variable name mistakes
(#244095)
Fixes: #242930
Assisted-by: Claude Opus 5.5
---
clang/lib/Headers/avx10_2_512minmaxintrin.h | 2 +-
clang/lib/Headers/avx10_2minmaxintrin.h | 2 +-
.../CodeGen/X86/avx10_2_512minmax-builtins.c | 19 ++++-
.../test/CodeGen/X86/avx10_2minmax-builtins.c | 72 +++++++++----------
4 files changed, 54 insertions(+), 41 deletions(-)
diff --git a/clang/lib/Headers/avx10_2_512minmaxintrin.h b/clang/lib/Headers/avx10_2_512minmaxintrin.h
index fbc7fbadbc6b2d..361c2c21ccd515 100644
--- a/clang/lib/Headers/avx10_2_512minmaxintrin.h
+++ b/clang/lib/Headers/avx10_2_512minmaxintrin.h
@@ -16,7 +16,7 @@
#define _mm512_minmax_pbh(A, B, C) \
((__m512bh)__builtin_ia32_vminmaxbf16512((__v32bf)(__m512bh)(A), \
- (__v32bf)(__m512bh)(A), (int)(C)))
+ (__v32bf)(__m512bh)(B), (int)(C)))
#define _mm512_mask_minmax_pbh(W, U, A, B, C) \
((__m512bh)__builtin_ia32_selectpbf_512( \
diff --git a/clang/lib/Headers/avx10_2minmaxintrin.h b/clang/lib/Headers/avx10_2minmaxintrin.h
index 809a01b04f13ef..cd7aab024ece63 100644
--- a/clang/lib/Headers/avx10_2minmaxintrin.h
+++ b/clang/lib/Headers/avx10_2minmaxintrin.h
@@ -88,7 +88,7 @@
#define _mm_mask_minmax_ph(W, U, A, B, C) \
((__m128h)__builtin_ia32_vminmaxph128_mask( \
(__v8hf)(__m128h)(A), (__v8hf)(__m128h)(B), (int)(C), \
- (__v8hf)(__m128h)(W), (__mmask16)-1))
+ (__v8hf)(__m128h)(W), (__mmask8)(U)))
#define _mm_maskz_minmax_ph(U, A, B, C) \
((__m128h)__builtin_ia32_vminmaxph128_mask( \
diff --git a/clang/test/CodeGen/X86/avx10_2_512minmax-builtins.c b/clang/test/CodeGen/X86/avx10_2_512minmax-builtins.c
index cbe4b634f6c5c5..f7f27ed0bbf738 100644
--- a/clang/test/CodeGen/X86/avx10_2_512minmax-builtins.c
+++ b/clang/test/CodeGen/X86/avx10_2_512minmax-builtins.c
@@ -9,20 +9,33 @@
__m512bh test_mm512_minmax_pbh(__m512bh __A, __m512bh __B) {
// CHECK-LABEL: @test_mm512_minmax_pbh(
- // CHECK: call <32 x bfloat> @llvm.x86.avx10.vminmaxbf16512(
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[AADDR:%[^,]+]],
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[BADDR:%[^,]+]],
+ // CHECK: [[X:%.+]] = load <32 x bfloat>, ptr [[AADDR]],
+ // CHECK: [[Y:%.+]] = load <32 x bfloat>, ptr [[BADDR]],
+ // CHECK: call <32 x bfloat> @llvm.x86.avx10.vminmaxbf16512(<32 x bfloat> [[X]], <32 x bfloat> [[Y]], i32 127)
return _mm512_minmax_pbh(__A, __B, 127);
}
__m512bh test_mm512_mask_minmax_pbh(__m512bh __A, __mmask32 __B, __m512bh __C, __m512bh __D) {
// CHECK-LABEL: @test_mm512_mask_minmax_pbh(
- // CHECK: call <32 x bfloat> @llvm.x86.avx10.vminmaxbf16512(
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[AADDR:%[^,]+]],
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[CADDR:%[^,]+]],
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[DADDR:%[^,]+]],
+ // CHECK: [[X:%.+]] = load <32 x bfloat>, ptr [[CADDR]],
+ // CHECK: [[Y:%.+]] = load <32 x bfloat>, ptr [[DADDR]],
+ // CHECK: call <32 x bfloat> @llvm.x86.avx10.vminmaxbf16512(<32 x bfloat> [[X]], <32 x bfloat> [[Y]], i32 127)
// CHECK: select <32 x i1> %{{.*}}, <32 x bfloat> %{{.*}}, <32 x bfloat> %{{.*}}
return _mm512_mask_minmax_pbh(__A, __B, __C, __D, 127);
}
__m512bh test_mm512_maskz_minmax_pbh(__mmask32 __A, __m512bh __B, __m512bh __C) {
// CHECK-LABEL: @test_mm512_maskz_minmax_pbh(
- // CHECK: call <32 x bfloat> @llvm.x86.avx10.vminmaxbf16512(
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[BADDR:%[^,]+]],
+ // CHECK: store <32 x bfloat> %{{.*}}, ptr [[CADDR:%[^,]+]],
+ // CHECK: [[X:%.+]] = load <32 x bfloat>, ptr [[BADDR]],
+ // CHECK: [[Y:%.+]] = load <32 x bfloat>, ptr [[CADDR]],
+ // CHECK: call <32 x bfloat> @llvm.x86.avx10.vminmaxbf16512(<32 x bfloat> [[X]], <32 x bfloat> [[Y]], i32 127)
// CHECK: zeroinitializer
// CHECK: select <32 x i1> %{{.*}}, <32 x bfloat> %{{.*}}, <32 x bfloat> %{{.*}}
return _mm512_maskz_minmax_pbh(__A, __B, __C, 127);
diff --git a/clang/test/CodeGen/X86/avx10_2minmax-builtins.c b/clang/test/CodeGen/X86/avx10_2minmax-builtins.c
index cce6d51c09969b..5c874de0edbc9d 100644
--- a/clang/test/CodeGen/X86/avx10_2minmax-builtins.c
+++ b/clang/test/CodeGen/X86/avx10_2minmax-builtins.c
@@ -51,216 +51,216 @@ __m256bh test_mm256_maskz_minmax_pbh(__mmask16 __A, __m256bh __B, __m256bh __C)
__m128d test_mm_minmax_pd(__m128d __A, __m128d __B) {
// CHECK-LABEL: @test_mm_minmax_pd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxpd128(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxpd128({{.*}}, i8 -1)
return _mm_minmax_pd(__A, __B, 127);
}
__m128d test_mm_mask_minmax_pd(__m128d __A, __mmask8 __B, __m128d __C, __m128d __D) {
// CHECK-LABEL: @test_mm_mask_minmax_pd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxpd128(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxpd128({{.*}}, i8 %{{.*}})
return _mm_mask_minmax_pd(__A, __B, __C, __D, 127);
}
__m128d test_mm_maskz_minmax_pd(__mmask8 __A, __m128d __B, __m128d __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_pd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxpd128(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxpd128({{.*}}, i8 %{{.*}})
return _mm_maskz_minmax_pd(__A, __B, __C, 127);
}
__m256d test_mm256_minmax_pd(__m256d __A, __m256d __B) {
// CHECK-LABEL: @test_mm256_minmax_pd(
- // CHECK: call <4 x double> @llvm.x86.avx10.mask.vminmaxpd256(
+ // CHECK: call <4 x double> @llvm.x86.avx10.mask.vminmaxpd256({{.*}}, i8 -1)
return _mm256_minmax_pd(__A, __B, 127);
}
__m256d test_mm256_mask_minmax_pd(__m256d __A, __mmask8 __B, __m256d __C, __m256d __D) {
// CHECK-LABEL: @test_mm256_mask_minmax_pd(
- // CHECK: call <4 x double> @llvm.x86.avx10.mask.vminmaxpd256(
+ // CHECK: call <4 x double> @llvm.x86.avx10.mask.vminmaxpd256({{.*}}, i8 %{{.*}})
return _mm256_mask_minmax_pd(__A, __B, __C, __D, 127);
}
__m256d test_mm256_maskz_minmax_pd(__mmask8 __A, __m256d __B, __m256d __C) {
// CHECK-LABEL: @test_mm256_maskz_minmax_pd(
- // CHECK: call <4 x double> @llvm.x86.avx10.mask.vminmaxpd256(
+ // CHECK: call <4 x double> @llvm.x86.avx10.mask.vminmaxpd256({{.*}}, i8 %{{.*}})
return _mm256_maskz_minmax_pd(__A, __B, __C, 127);
}
__m128h test_mm_minmax_ph(__m128h __A, __m128h __B) {
// CHECK-LABEL: @test_mm_minmax_ph(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxph128(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxph128({{.*}}, i8 -1)
return _mm_minmax_ph(__A, __B, 127);
}
__m128h test_mm_mask_minmax_ph(__m128h __A, __mmask8 __B, __m128h __C, __m128h __D) {
// CHECK-LABEL: @test_mm_mask_minmax_ph(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxph128(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxph128({{.*}}, i8 %{{.*}})
return _mm_mask_minmax_ph(__A, __B, __C, __D, 127);
}
__m128h test_mm_maskz_minmax_ph(__mmask8 __A, __m128h __B, __m128h __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_ph(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxph128(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxph128({{.*}}, i8 %{{.*}})
return _mm_maskz_minmax_ph(__A, __B, __C, 127);
}
__m256h test_mm256_minmax_ph(__m256h __A, __m256h __B) {
// CHECK-LABEL: @test_mm256_minmax_ph(
- // CHECK: call <16 x half> @llvm.x86.avx10.mask.vminmaxph256(
+ // CHECK: call <16 x half> @llvm.x86.avx10.mask.vminmaxph256({{.*}}, i16 -1)
return _mm256_minmax_ph(__A, __B, 127);
}
__m256h test_mm256_mask_minmax_ph(__m256h __A, __mmask16 __B, __m256h __C, __m256h __D) {
// CHECK-LABEL: @test_mm256_mask_minmax_ph(
- // CHECK: call <16 x half> @llvm.x86.avx10.mask.vminmaxph256(
+ // CHECK: call <16 x half> @llvm.x86.avx10.mask.vminmaxph256({{.*}}, i16 %{{.*}})
return _mm256_mask_minmax_ph(__A, __B, __C, __D, 127);
}
__m256h test_mm256_maskz_minmax_ph(__mmask16 __A, __m256h __B, __m256h __C) {
// CHECK-LABEL: @test_mm256_maskz_minmax_ph(
- // CHECK: call <16 x half> @llvm.x86.avx10.mask.vminmaxph256(
+ // CHECK: call <16 x half> @llvm.x86.avx10.mask.vminmaxph256({{.*}}, i16 %{{.*}})
return _mm256_maskz_minmax_ph(__A, __B, __C, 127);
}
__m128 test_mm_minmax_ps(__m128 __A, __m128 __B) {
// CHECK-LABEL: @test_mm_minmax_ps(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxps128(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxps128({{.*}}, i8 -1)
return _mm_minmax_ps(__A, __B, 127);
}
__m128 test_mm_mask_minmax_ps(__m128 __A, __mmask8 __B, __m128 __C, __m128 __D) {
// CHECK-LABEL: @test_mm_mask_minmax_ps(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxps128(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxps128({{.*}}, i8 %{{.*}})
return _mm_mask_minmax_ps(__A, __B, __C, __D, 127);
}
__m128 test_mm_maskz_minmax_ps(__mmask8 __A, __m128 __B, __m128 __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_ps(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxps128(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxps128({{.*}}, i8 %{{.*}})
return _mm_maskz_minmax_ps(__A, __B, __C, 127);
}
__m256 test_mm256_minmax_ps(__m256 __A, __m256 __B) {
// CHECK-LABEL: @test_mm256_minmax_ps(
- // CHECK: call <8 x float> @llvm.x86.avx10.mask.vminmaxps256(
+ // CHECK: call <8 x float> @llvm.x86.avx10.mask.vminmaxps256({{.*}}, i8 -1)
return _mm256_minmax_ps(__A, __B, 127);
}
__m256 test_mm256_mask_minmax_ps(__m256 __A, __mmask8 __B, __m256 __C, __m256 __D) {
// CHECK-LABEL: @test_mm256_mask_minmax_ps(
- // CHECK: call <8 x float> @llvm.x86.avx10.mask.vminmaxps256(
+ // CHECK: call <8 x float> @llvm.x86.avx10.mask.vminmaxps256({{.*}}, i8 %{{.*}})
return _mm256_mask_minmax_ps(__A, __B, __C, __D, 127);
}
__m256 test_mm256_maskz_minmax_ps(__mmask8 __A, __m256 __B, __m256 __C) {
// CHECK-LABEL: @test_mm256_maskz_minmax_ps(
- // CHECK: call <8 x float> @llvm.x86.avx10.mask.vminmaxps256(
+ // CHECK: call <8 x float> @llvm.x86.avx10.mask.vminmaxps256({{.*}}, i8 %{{.*}})
return _mm256_maskz_minmax_ps(__A, __B, __C, 127);
}
__m128d test_mm_minmax_sd(__m128d __A, __m128d __B) {
// CHECK-LABEL: @test_mm_minmax_sd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round({{.*}}, i8 -1, i32 {{.*}})
return _mm_minmax_sd(__A, __B, 127);
}
__m128d test_mm_mask_minmax_sd(__m128d __A, __mmask8 __B, __m128d __C, __m128d __D) {
// CHECK-LABEL: @test_mm_mask_minmax_sd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_mask_minmax_sd(__A, __B, __C, __D, 127);
}
__m128d test_mm_maskz_minmax_sd(__mmask8 __A, __m128d __B, __m128d __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_sd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_maskz_minmax_sd(__A, __B, __C, 127);
}
__m128d test_mm_minmax_round_sd(__m128d __A, __m128d __B) {
// CHECK-LABEL: @test_mm_minmax_round_sd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round({{.*}}, i8 -1, i32 {{.*}})
return _mm_minmax_round_sd(__A, __B, 127, _MM_FROUND_NO_EXC);
}
__m128d test_mm_mask_minmax_round_sd(__m128d __A, __mmask8 __B, __m128d __C, __m128d __D) {
// CHECK-LABEL: @test_mm_mask_minmax_round_sd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_mask_minmax_round_sd(__A, __B, __C, __D, 127, _MM_FROUND_NO_EXC);
}
__m128d test_mm_maskz_minmax_round_sd(__mmask8 __A, __m128d __B, __m128d __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_round_sd(
- // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round(
+ // CHECK: call <2 x double> @llvm.x86.avx10.mask.vminmaxsd.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_maskz_minmax_round_sd(__A, __B, __C, 127, _MM_FROUND_NO_EXC);
}
__m128h test_mm_minmax_sh(__m128h __A, __m128h __B) {
// CHECK-LABEL: @test_mm_minmax_sh(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round({{.*}}, i8 -1, i32 {{.*}})
return _mm_minmax_sh(__A, __B, 127);
}
__m128h test_mm_mask_minmax_sh(__m128h __A, __mmask8 __B, __m128h __C, __m128h __D) {
// CHECK-LABEL: @test_mm_mask_minmax_sh(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_mask_minmax_sh(__A, __B, __C, __D, 127);
}
__m128h test_mm_maskz_minmax_sh(__mmask8 __A, __m128h __B, __m128h __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_sh(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_maskz_minmax_sh(__A, __B, __C, 127);
}
__m128h test_mm_minmax_round_sh(__m128h __A, __m128h __B) {
// CHECK-LABEL: @test_mm_minmax_round_sh(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round({{.*}}, i8 -1, i32 {{.*}})
return _mm_minmax_round_sh(__A, __B, 127, _MM_FROUND_NO_EXC);
}
__m128h test_mm_mask_minmax_round_sh(__m128h __A, __mmask8 __B, __m128h __C, __m128h __D) {
// CHECK-LABEL: @test_mm_mask_minmax_round_sh(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_mask_minmax_round_sh(__A, __B, __C, __D, 127, _MM_FROUND_NO_EXC);
}
__m128h test_mm_maskz_minmax_round_sh(__mmask8 __A, __m128h __B, __m128h __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_round_sh(
- // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round(
+ // CHECK: call <8 x half> @llvm.x86.avx10.mask.vminmaxsh.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_maskz_minmax_round_sh(__A, __B, __C, 127, _MM_FROUND_NO_EXC);
}
__m128 test_mm_minmax_ss(__m128 __A, __m128 __B) {
// CHECK-LABEL: @test_mm_minmax_ss(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round({{.*}}, i8 -1, i32 {{.*}})
return _mm_minmax_ss(__A, __B, 127);
}
__m128 test_mm_mask_minmax_ss(__m128 __A, __mmask8 __B, __m128 __C, __m128 __D) {
// CHECK-LABEL: @test_mm_mask_minmax_ss(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_mask_minmax_ss(__A, __B, __C, __D, 127);
}
__m128 test_mm_maskz_minmax_ss(__mmask8 __A, __m128 __B, __m128 __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_ss(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_maskz_minmax_ss(__A, __B, __C, 127);
}
__m128 test_mm_minmax_round_ss(__m128 __A, __m128 __B) {
// CHECK-LABEL: @test_mm_minmax_round_ss(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round({{.*}}, i8 -1, i32 {{.*}})
return _mm_minmax_round_ss(__A, __B, 127, _MM_FROUND_NO_EXC);
}
__m128 test_mm_mask_minmax_round_ss(__m128 __A, __mmask8 __B, __m128 __C, __m128 __D) {
// CHECK-LABEL: @test_mm_mask_minmax_round_ss(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_mask_minmax_round_ss(__A, __B, __C, __D, 127, _MM_FROUND_NO_EXC);
}
__m128 test_mm_maskz_minmax_round_ss(__mmask8 __A, __m128 __B, __m128 __C) {
// CHECK-LABEL: @test_mm_maskz_minmax_round_ss(
- // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round(
+ // CHECK: call <4 x float> @llvm.x86.avx10.mask.vminmaxss.round({{.*}}, i8 %{{.*}}, i32 {{.*}})
return _mm_maskz_minmax_round_ss(__A, __B, __C, 127, _MM_FROUND_NO_EXC);
}
>From c9befbbe32a9d1a5a76237663277235f87aec596 Mon Sep 17 00:00:00 2001
From: Sergei Barannikov <barannikov88 at gmail.com>
Date: Sat, 21 Nov 2026 16:38:48 +0300
Subject: [PATCH 22/35] [clang] Correct the argument declaration of
'ptrauth_vtable_pointer' in Attrs.td (#243446)
The three enum parameters of 'ptrauth_vtable_pointer' were declared with
IsString set, but its custom handler requires *identifier* arguments in
those positions and rejects anything else.
There is no observable behavioral difference since the attribute does
not take the generic path where IsString would matter.
Found by observation.
---
clang/include/clang/Basic/Attr.td | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/clang/include/clang/Basic/Attr.td b/clang/include/clang/Basic/Attr.td
index d68e3253628d8d..35afd9d197cfde 100644
--- a/clang/include/clang/Basic/Attr.td
+++ b/clang/include/clang/Basic/Attr.td
@@ -5517,19 +5517,19 @@ def VTablePointerAuthentication : InheritableAttr {
let Subjects = SubjectList<[CXXRecord]>;
let Documentation = [Undocumented];
let StrictEnumParameters = 1;
- let Args = [EnumArgument<"Key", "VPtrAuthKeyType", /*is_string=*/ true,
+ let Args = [EnumArgument<"Key", "VPtrAuthKeyType", /*is_string=*/ false,
["default_key", "no_authentication", "process_dependent",
"process_independent"],
["DefaultKey", "NoKey", "ProcessDependent",
"ProcessIndependent"]>,
EnumArgument<"AddressDiscrimination", "AddressDiscriminationMode",
- /*is_string=*/ true,
+ /*is_string=*/ false,
["default_address_discrimination", "no_address_discrimination",
"address_discrimination"],
["DefaultAddressDiscrimination", "NoAddressDiscrimination",
"AddressDiscrimination"]>,
EnumArgument<"ExtraDiscrimination", "ExtraDiscrimination",
- /*is_string=*/ true,
+ /*is_string=*/ false,
["default_extra_discrimination", "no_extra_discrimination",
"type_discrimination", "custom_discrimination"],
["DefaultExtraDiscrimination", "NoExtraDiscrimination",
>From a33a508a152f832bd080f455d5ac8ec44cd6cf57 Mon Sep 17 00:00:00 2001
From: Alan Li <me at alanli.org>
Date: Sat, 21 Nov 2026 06:06:54 -0800
Subject: [PATCH 23/35] [CodeGen] Fix physical-register input latency for
inline asm (#229764)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
As described in #229538, the scheduling DAG classifies inline asm’s
variable inputs as pseudo uses, assigning zero latency and hiding
modeled producer latency from scheduling decisions. The dependency
preserves instruction order, but the scheduler does not account for the
time needed to produce the result.
This patch is a more generic approach which fix the issue in general
infrastructure.
Fix approach:
* In `ScheduleDAGInstrs`, recognize explicit inline-asm inputs as real
uses and allow explicit definitions feeding them to bypass the
pseudo-definition check.
* The existing `computeOperandLatency()` then supplies modeled producer
latency, including for asm-to-asm dependencies.
* Implicit bookkeeping operands and other consumers retain their
existing behavior.
Also addded AMDGPU and X86 MIR coverage for producer latency, asm-to-asm
dependencies, and bookkeeping operands.
---
llvm/lib/CodeGen/ScheduleDAGInstrs.cpp | 13 +-
llvm/test/CodeGen/AArch64/arm64-inline-asm.ll | 2 +-
llvm/test/CodeGen/AArch64/preserve_nonecc.ll | 10 +-
.../AArch64/speculation-hardening-loads.ll | 2 +-
.../CodeGen/AArch64/strictfp-inlineasm.ll | 2 +-
.../test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll | 80 +++---
.../CodeGen/AMDGPU/a-v-global-atomicrmw.ll | 4 +-
llvm/test/CodeGen/AMDGPU/add.ll | 10 +-
llvm/test/CodeGen/AMDGPU/agpr-remat.ll | 2 +-
.../CodeGen/AMDGPU/captured-frame-index.ll | 6 +-
llvm/test/CodeGen/AMDGPU/fneg-combines.ll | 4 +-
llvm/test/CodeGen/AMDGPU/gds-allocation.ll | 2 +-
.../gfx-callable-preserved-registers.ll | 22 +-
.../AMDGPU/inline-asm-vgpr-sgpr-copy.ll | 4 +-
.../AMDGPU/inlineasm-ds-return-latency.mir | 47 ++++
.../AMDGPU/inlineasm-producer-latency.mir | 47 ++++
.../CodeGen/AMDGPU/insert_vector_elt.v2i16.ll | 6 +-
.../AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll | 24 +-
.../AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll | 4 +-
.../AMDGPU/llvm.amdgcn.readfirstlane.ll | 84 +++---
.../materialize-frame-index-sgpr.gfx10.ll | 54 ++--
.../AMDGPU/materialize-frame-index-sgpr.ll | 4 +-
.../AMDGPU/postra-coexec-inlineasm-flags.mir | 48 ++++
.../postra-coexec-inlineasm-latency.mir | 93 +++++++
.../AMDGPU/postra-inlineasm-def-latency.mir | 26 ++
.../AMDGPU/preserve-wwm-copy-dst-reg.ll | 72 +++---
llvm/test/CodeGen/AMDGPU/s_cmp_0.ll | 2 +-
.../scc-clobbered-sgpr-to-vmem-spill.ll | 22 +-
.../AMDGPU/shufflevector.v2i64.v8i64.ll | 242 +++++++++---------
llvm/test/CodeGen/AMDGPU/spill-agpr.ll | 14 +-
llvm/test/CodeGen/AMDGPU/sub.ll | 10 +-
.../AMDGPU/undefined-subreg-liverange.ll | 4 +-
.../CodeGen/AMDGPU/wmma-scale-src-perm.ll | 6 +-
llvm/test/CodeGen/AMDGPU/write_register.ll | 4 +-
.../PowerPC/inline-asm-physical-fpr-spe.ll | 2 +-
.../CodeGen/PowerPC/not-fixed-frame-object.ll | 4 +-
llvm/test/CodeGen/RISCV/load-store-pair.ll | 20 +-
.../RISCV/rvv/vsetvl-cross-inline-asm.ll | 2 +-
.../CodeGen/RISCV/tlsdesc-clobber-vector.ll | 80 +++---
...inline-asm-fp-int-casting-explicit-regs.ll | 4 +-
.../SystemZ/inline-asm-fp-int-casting.ll | 2 +-
.../Thumb2/LowOverheadLoops/inlineasm.ll | 8 +-
.../Thumb2/arm_canberra_distance_f32.ll | 4 +-
.../CodeGen/X86/inlineasm-physreg-latency.mir | 173 +++++++++++++
44 files changed, 859 insertions(+), 416 deletions(-)
create mode 100644 llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir
create mode 100644 llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir
create mode 100644 llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir
create mode 100644 llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir
create mode 100644 llvm/test/CodeGen/AMDGPU/postra-inlineasm-def-latency.mir
create mode 100644 llvm/test/CodeGen/X86/inlineasm-physreg-latency.mir
diff --git a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
index 420152f1a09ab8..6e7d56f1ff9a51 100644
--- a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
+++ b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
@@ -206,6 +206,8 @@ void ScheduleDAGInstrs::addPhysRegDataDeps(SUnit *SU, unsigned OperIdx) {
const MCInstrDesc &DefMIDesc = SU->getInstr()->getDesc();
bool ImplicitPseudoDef = (OperIdx >= DefMIDesc.getNumOperands() &&
!DefMIDesc.hasImplicitDefOfPhysReg(Reg));
+ bool IsInlineAsmDef = SU->getInstr()->isInlineAsm() &&
+ SU->getInstr()->findInlineAsmFlagIdx(OperIdx) >= 0;
for (MCRegUnit Unit : TRI->regunits(Reg)) {
for (RegUnit2SUnitsMap::iterator I = Uses.find(Unit); I != Uses.end();
++I) {
@@ -218,6 +220,7 @@ void ScheduleDAGInstrs::addPhysRegDataDeps(SUnit *SU, unsigned OperIdx) {
MachineInstr *UseInstr = nullptr;
int UseOpIdx = I->OpIdx;
bool ImplicitPseudoUse = false;
+ bool IsInlineAsmUse = false;
SDep Dep;
if (UseOpIdx < 0) {
Dep = SDep(SU, SDep::Artificial);
@@ -229,12 +232,18 @@ void ScheduleDAGInstrs::addPhysRegDataDeps(SUnit *SU, unsigned OperIdx) {
UseInstr = UseSU->getInstr();
Register UseReg = UseInstr->getOperand(UseOpIdx).getReg();
const MCInstrDesc &UseMIDesc = UseInstr->getDesc();
+ // Recognize asm inputs by their operand groups, excluding trailing
+ // register bookkeeping operands.
+ IsInlineAsmUse = UseInstr->isInlineAsm() &&
+ UseInstr->findInlineAsmFlagIdx(UseOpIdx) >= 0;
ImplicitPseudoUse = UseOpIdx >= ((int)UseMIDesc.getNumOperands()) &&
- !UseMIDesc.hasImplicitUseOfPhysReg(UseReg);
+ !UseMIDesc.hasImplicitUseOfPhysReg(UseReg) &&
+ !IsInlineAsmUse;
Dep = SDep(SU, SDep::Data, UseReg);
}
- if (!ImplicitPseudoDef && !ImplicitPseudoUse) {
+ if ((!ImplicitPseudoDef && !ImplicitPseudoUse) ||
+ (IsInlineAsmDef && IsInlineAsmUse)) {
Dep.setLatency(SchedModel.computeOperandLatency(SU->getInstr(), OperIdx,
UseInstr, UseOpIdx));
} else {
diff --git a/llvm/test/CodeGen/AArch64/arm64-inline-asm.ll b/llvm/test/CodeGen/AArch64/arm64-inline-asm.ll
index 8947563e31f855..f2e801decf4bb1 100644
--- a/llvm/test/CodeGen/AArch64/arm64-inline-asm.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-inline-asm.ll
@@ -367,13 +367,13 @@ entry:
define void @test_zero_reg(ptr %addr) {
; CHECK-LABEL: test_zero_reg:
; CHECK: ; %bb.0:
+; CHECK-NEXT: mov w8, #1 ; =0x1
; CHECK-NEXT: ; InlineAsm Start
; CHECK-NEXT: USE(xzr)
; CHECK-NEXT: ; InlineAsm End
; CHECK-NEXT: ; InlineAsm Start
; CHECK-NEXT: USE(wzr)
; CHECK-NEXT: ; InlineAsm End
-; CHECK-NEXT: mov w8, #1 ; =0x1
; CHECK-NEXT: ; InlineAsm Start
; CHECK-NEXT: USE(w8)
; CHECK-NEXT: ; InlineAsm End
diff --git a/llvm/test/CodeGen/AArch64/preserve_nonecc.ll b/llvm/test/CodeGen/AArch64/preserve_nonecc.ll
index 2ef9f650d70fb0..c1674913ebadb6 100644
--- a/llvm/test/CodeGen/AArch64/preserve_nonecc.ll
+++ b/llvm/test/CodeGen/AArch64/preserve_nonecc.ll
@@ -38,6 +38,11 @@ entry:
;ALL: stp d12, d11
;ALL: stp d14, d13
;ALL: stp d16, d15
+;ALL: ldp x9, x8
+;ALL: ldp x11, x10
+;ALL: ldp x13, x12
+;ALL: ldp x15, x14
+;ALL: ldp x17, x16
;ALL: ldp x20, x19
;ALL: ldp x22, x21
;ALL: ldp x24, x23
@@ -48,11 +53,6 @@ entry:
;ALL: ldp d12, d11
;ALL: ldp d14, d13
;ALL: ldp d16, d15
-;ALL: ldp x9, x8
-;ALL: ldp x11, x10
-;ALL: ldp x13, x12
-;ALL: ldp x15, x14
-;ALL: ldp x17, x16
;ALL: InlineAsm Start
%a0 = call i64 asm sideeffect "", "={x8}"() nounwind
%a1 = call i64 asm sideeffect "", "={x9}"() nounwind
diff --git a/llvm/test/CodeGen/AArch64/speculation-hardening-loads.ll b/llvm/test/CodeGen/AArch64/speculation-hardening-loads.ll
index ef9309959c5135..3e8ad31677f211 100644
--- a/llvm/test/CodeGen/AArch64/speculation-hardening-loads.ll
+++ b/llvm/test/CodeGen/AArch64/speculation-hardening-loads.ll
@@ -96,8 +96,8 @@ define i64 @no_masking_with_full_control_flow_barriers(i64 %a, i64 %b, ptr %p) s
; CHECK-NEXT: dsb sy
; CHECK-NEXT: isb
; CHECK-NEXT: ldr x8, [x2]
-; CHECK-NEXT: mov x17, x0
; CHECK-NEXT: mov x16, x1
+; CHECK-NEXT: mov x17, x0
; CHECK-NEXT: //APP
; CHECK-NEXT: hint #12
; CHECK-NEXT: //NO_APP
diff --git a/llvm/test/CodeGen/AArch64/strictfp-inlineasm.ll b/llvm/test/CodeGen/AArch64/strictfp-inlineasm.ll
index 0bbf31c5c0d738..d2fe8d0b7fafe9 100644
--- a/llvm/test/CodeGen/AArch64/strictfp-inlineasm.ll
+++ b/llvm/test/CodeGen/AArch64/strictfp-inlineasm.ll
@@ -4,8 +4,8 @@
define i32 @foo() strictfp {
; CHECK-LABEL: foo:
; CHECK: // %bb.0: // %entry
-; CHECK-NEXT: mov w0, #-1 // =0xffffffff
; CHECK-NEXT: mov w8, #1 // =0x1
+; CHECK-NEXT: mov w0, #-1 // =0xffffffff
; CHECK-NEXT: //APP
; CHECK-NEXT: //NO_APP
; CHECK-NEXT: ret
diff --git a/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll b/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll
index 6525309e8d42fd..0e80dc46dc34d6 100644
--- a/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll
+++ b/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll
@@ -536,8 +536,8 @@ define void @flat_atomic_xchg_i32_ret_av_av_no_agprs(ptr %ptr) #0 {
; GFX950-NEXT: scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v23, a27 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v22, a28 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v21, a29 ; Reload Reuse
@@ -4289,8 +4289,8 @@ define void @flat_atomic_xor_i32_ret_av_av_no_agprs(ptr %ptr) #0 {
; GFX950-NEXT: scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v23, a27 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v22, a28 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v21, a29 ; Reload Reuse
@@ -14665,11 +14665,11 @@ define void @flat_atomic_add_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(1)
; GFX90A-NEXT: v_add_co_u32_e32 v0, vcc, v2, v0
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_addc_co_u32_e32 v1, vcc, v3, v1, vcc
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -14705,9 +14705,9 @@ define void @flat_atomic_add_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: scratch_load_dwordx2 v[2:3], off, s0
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_lshl_add_u64 v[0:1], v[2:3], 0, v[0:1]
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -14828,11 +14828,11 @@ define void @flat_atomic_sub_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(1)
; GFX90A-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v0
; GFX90A-NEXT: s_waitcnt vmcnt(0)
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v1, vcc
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -14871,8 +14871,8 @@ define void @flat_atomic_sub_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: s_nop 0
; GFX950-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v1, vcc
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -14996,10 +14996,10 @@ define void @flat_atomic_and_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: v_and_b32_e32 v1, v3, v1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_and_b32_e32 v0, v2, v0
-; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -15036,9 +15036,9 @@ define void @flat_atomic_and_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_and_b32_e32 v1, v3, v1
; GFX950-NEXT: v_and_b32_e32 v0, v2, v0
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -15390,10 +15390,10 @@ define void @flat_atomic_or_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: v_or_b32_e32 v1, v3, v1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -15430,9 +15430,9 @@ define void @flat_atomic_or_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_or_b32_e32 v1, v3, v1
; GFX950-NEXT: v_or_b32_e32 v0, v2, v0
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -15555,10 +15555,10 @@ define void @flat_atomic_xor_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: v_xor_b32_e32 v1, v3, v1
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_xor_b32_e32 v0, v2, v0
-; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -15595,9 +15595,9 @@ define void @flat_atomic_xor_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_xor_b32_e32 v1, v3, v1
; GFX950-NEXT: v_xor_b32_e32 v0, v2, v0
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -15719,11 +15719,11 @@ define void @flat_atomic_max_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_gt_i64_e32 vcc, v[2:3], v[0:1]
; GFX90A-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -15888,11 +15888,11 @@ define void @flat_atomic_min_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_le_i64_e32 vcc, v[2:3], v[0:1]
; GFX90A-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -16057,11 +16057,11 @@ define void @flat_atomic_umax_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1]
; GFX90A-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -16226,11 +16226,11 @@ define void @flat_atomic_umin_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_cmp_le_u64_e32 vcc, v[2:3], v[0:1]
; GFX90A-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -16398,11 +16398,11 @@ define void @flat_atomic_uinc_wrap_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: v_addc_co_u32_e32 v6, vcc, 0, v3, vcc
; GFX90A-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1]
; GFX90A-NEXT: v_cndmask_b32_e32 v1, 0, v5, vcc
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_cndmask_b32_e32 v0, 0, v6, vcc
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen offset:4
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -16575,11 +16575,11 @@ define void @flat_atomic_udec_wrap_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: v_cmp_gt_u64_e64 s[4:5], v[2:3], v[0:1]
; GFX90A-NEXT: s_or_b64 vcc, vcc, s[4:5]
; GFX90A-NEXT: v_cndmask_b32_e32 v1, v6, v1, vcc
+; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: v_cndmask_b32_e32 v0, v5, v0, vcc
; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
-; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -16622,9 +16622,9 @@ define void @flat_atomic_udec_wrap_i64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: s_nop 0
; GFX950-NEXT: v_cndmask_b32_e32 v1, v5, v1, vcc
; GFX950-NEXT: v_cndmask_b32_e32 v0, v4, v0, vcc
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s2
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s2
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -17250,8 +17250,8 @@ define void @flat_atomic_fadd_f32_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: buffer_load_dword v2, v1, s[0:3], 0 offen
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_add_f32_e32 v0, v2, v0
-; GFX90A-NEXT: buffer_store_dword v0, v1, s[0:3], 0 offen
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
+; GFX90A-NEXT: buffer_store_dword v0, v1, s[0:3], 0 offen
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a0
; GFX90A-NEXT: ;;#ASMEND
@@ -18080,10 +18080,10 @@ define void @flat_atomic_fadd_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: buffer_load_dword v3, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_add_f64 v[0:1], v[2:3], v[0:1]
-; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v4, s[0:3], 0 offen offset:4
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -18131,9 +18131,9 @@ define void @flat_atomic_fadd_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: scratch_load_dwordx2 v[2:3], off, s0
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_add_f64 v[0:1], v[2:3], v[0:1]
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -18489,10 +18489,10 @@ define void @flat_atomic_fmax_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
; GFX90A-NEXT: v_max_f64 v[0:1], v[4:5], v[0:1]
-; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -18530,9 +18530,9 @@ define void @flat_atomic_fmax_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
; GFX950-NEXT: v_max_f64 v[0:1], v[4:5], v[0:1]
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
@@ -18657,10 +18657,10 @@ define void @flat_atomic_fmin_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX90A-NEXT: s_waitcnt vmcnt(0)
; GFX90A-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
; GFX90A-NEXT: v_min_f64 v[0:1], v[4:5], v[0:1]
-; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen
-; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4
; GFX90A-NEXT: v_accvgpr_write_b32 a0, v2
; GFX90A-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen
+; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use a[0:1]
; GFX90A-NEXT: ;;#ASMEND
@@ -18698,9 +18698,9 @@ define void @flat_atomic_fmin_f64_saddr_ret_a_a(ptr inreg %ptr) #0 {
; GFX950-NEXT: s_waitcnt vmcnt(0)
; GFX950-NEXT: v_max_f64 v[4:5], v[2:3], v[2:3]
; GFX950-NEXT: v_min_f64 v[0:1], v[4:5], v[0:1]
-; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: v_accvgpr_write_b32 a0, v2
; GFX950-NEXT: v_accvgpr_write_b32 a1, v3
+; GFX950-NEXT: scratch_store_dwordx2 off, v[0:1], s0
; GFX950-NEXT: ;;#ASMSTART
; GFX950-NEXT: ; use a[0:1]
; GFX950-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll b/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll
index 82632c0c94c32c..d8e955aa79f029 100644
--- a/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll
+++ b/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll
@@ -536,8 +536,8 @@ define void @global_atomic_xchg_i32_ret_av_av_no_agprs(ptr addrspace(1) %ptr) #0
; GFX950-NEXT: scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v23, a27 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v22, a28 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v21, a29 ; Reload Reuse
@@ -3262,8 +3262,8 @@ define void @global_atomic_xor_i32_ret_av_av_no_agprs(ptr addrspace(1) %ptr) #0
; GFX950-NEXT: scratch_load_dwordx4 v[4:7], off, s32 offset:16 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[8:11], off, s32 offset:32 ; 16-byte Folded Reload
; GFX950-NEXT: scratch_load_dwordx4 v[12:15], off, s32 offset:48 ; 16-byte Folded Reload
-; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: scratch_load_dwordx3 v[16:18], off, s32 offset:64 ; 12-byte Folded Reload
+; GFX950-NEXT: v_accvgpr_read_b32 v19, a31 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v23, a27 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v22, a28 ; Reload Reuse
; GFX950-NEXT: v_accvgpr_read_b32 v21, a29 ; Reload Reuse
diff --git a/llvm/test/CodeGen/AMDGPU/add.ll b/llvm/test/CodeGen/AMDGPU/add.ll
index 8124d8d8efc5dd..4623cd9cfad9ae 100644
--- a/llvm/test/CodeGen/AMDGPU/add.ll
+++ b/llvm/test/CodeGen/AMDGPU/add.ll
@@ -1353,11 +1353,11 @@ endif:
define amdgpu_ps void @add_select_vop3(i32 inreg %s, i32 %v) {
; GFX6-LABEL: add_select_vop3:
; GFX6: ; %bb.0:
-; GFX6-NEXT: v_add_i32_e64 v0, s[0:1], s0, v0
-; GFX6-NEXT: s_mov_b32 m0, -1
; GFX6-NEXT: ;;#ASMSTART
; GFX6-NEXT: ; def vcc
; GFX6-NEXT: ;;#ASMEND
+; GFX6-NEXT: v_add_i32_e64 v0, s[0:1], s0, v0
+; GFX6-NEXT: s_mov_b32 m0, -1
; GFX6-NEXT: ds_write_b32 v0, v0
; GFX6-NEXT: ;;#ASMSTART
; GFX6-NEXT: ; use vcc
@@ -1366,11 +1366,11 @@ define amdgpu_ps void @add_select_vop3(i32 inreg %s, i32 %v) {
;
; GFX8-LABEL: add_select_vop3:
; GFX8: ; %bb.0:
-; GFX8-NEXT: v_add_u32_e64 v0, s[0:1], s0, v0
-; GFX8-NEXT: s_mov_b32 m0, -1
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; def vcc
; GFX8-NEXT: ;;#ASMEND
+; GFX8-NEXT: v_add_u32_e64 v0, s[0:1], s0, v0
+; GFX8-NEXT: s_mov_b32 m0, -1
; GFX8-NEXT: ds_write_b32 v0, v0
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use vcc
@@ -1379,10 +1379,10 @@ define amdgpu_ps void @add_select_vop3(i32 inreg %s, i32 %v) {
;
; GFX9-LABEL: add_select_vop3:
; GFX9: ; %bb.0:
-; GFX9-NEXT: v_add_u32_e32 v0, s0, v0
; GFX9-NEXT: ;;#ASMSTART
; GFX9-NEXT: ; def vcc
; GFX9-NEXT: ;;#ASMEND
+; GFX9-NEXT: v_add_u32_e32 v0, s0, v0
; GFX9-NEXT: ds_write_b32 v0, v0
; GFX9-NEXT: ;;#ASMSTART
; GFX9-NEXT: ; use vcc
diff --git a/llvm/test/CodeGen/AMDGPU/agpr-remat.ll b/llvm/test/CodeGen/AMDGPU/agpr-remat.ll
index ede7ce390e1598..dd7506297fa0e4 100644
--- a/llvm/test/CodeGen/AMDGPU/agpr-remat.ll
+++ b/llvm/test/CodeGen/AMDGPU/agpr-remat.ll
@@ -32,9 +32,9 @@ define void @remat_regcopy_avoids_spill(i32 %v0, i32 %v1, i32 %v2, i32 %v3, i32
; GFX908-NEXT: v_accvgpr_write_b32 a2, v0
; GFX908-NEXT: v_accvgpr_write_b32 a3, v1
; GFX908-NEXT: v_accvgpr_write_b32 a4, v2
+; GFX908-NEXT: v_accvgpr_write_b32 a5, v3
; GFX908-NEXT: v_accvgpr_write_b32 a0, v7
; GFX908-NEXT: v_accvgpr_write_b32 a1, v8
-; GFX908-NEXT: v_accvgpr_write_b32 a5, v3
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ;;#ASMEND
; GFX908-NEXT: v_accvgpr_write_b32 a2, v4
diff --git a/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll b/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
index 28d650013fcec5..5916b08be4727b 100644
--- a/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
+++ b/llvm/test/CodeGen/AMDGPU/captured-frame-index.ll
@@ -273,10 +273,10 @@ define void @func_alloca_offset_use_asm_sgpr() {
; GCN: v_lshr_b32_e64 [[V_FI:v[0-9]+]], s32, 6
; GCN: s_movk_i32 vcc_lo, 0x4010
; GCN: s_add_i32 [[S_FI]], [[S_FI]], 16
+; GCN-NEXT: v_add_i32_e32 [[V_FI]], vcc, vcc_lo, [[V_FI]]
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: ; use [[S_FI]]
; GCN-NEXT: ;;#ASMEND
-; GCN-NEXT: v_add_i32_e32 [[V_FI:v[0-9]+]], vcc, vcc_lo, [[V_FI]]
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: ; use [[V_FI]]
; GCN-NEXT: ;;#ASMEND
@@ -295,7 +295,7 @@ define void @func_alloca_offset_use_asm_vgpr() {
; GCN-NEXT: ; use [[FI0]]
; GCN-NEXT: ;;#ASMEND
; GCN: s_movk_i32 [[FI1:s[0-9]+]], 0x4010
-; GCN-NEXT: ;;#ASMSTART
+; GCN: ;;#ASMSTART
; GCN-NEXT: ; use [[FI1]]
; GCN-NEXT: ;;#ASMEND
define amdgpu_kernel void @kernel_alloca_offset_use_asm_sgpr() {
@@ -314,7 +314,7 @@ define amdgpu_kernel void @kernel_alloca_offset_use_asm_sgpr() {
; GCN-NEXT: ;;#ASMEND
; GCN: v_mov_b32_e32 v0, 0x4010
-; GCN-NEXT: ;;#ASMSTART
+; GCN: ;;#ASMSTART
; GCN-NEXT: ; use v0
; GCN-NEXT: ;;#ASMEND
define amdgpu_kernel void @kernel_alloca_offset_use_asm_vgpr() {
diff --git a/llvm/test/CodeGen/AMDGPU/fneg-combines.ll b/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
index 5395bb784fc935..21de0e507c162d 100644
--- a/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
+++ b/llvm/test/CodeGen/AMDGPU/fneg-combines.ll
@@ -7203,11 +7203,11 @@ define amdgpu_kernel void @v_fneg_inlineasm_multi_use_src_f32(ptr addrspace(1) %
; SI-NEXT: v_mov_b32_e32 v3, s5
; SI-NEXT: v_add_i32_e32 v2, vcc, s4, v4
; SI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
-; SI-NEXT: v_mov_b32_e32 v5, s7
; SI-NEXT: flat_load_dword v6, v[0:1] glc
; SI-NEXT: s_waitcnt vmcnt(0)
; SI-NEXT: flat_load_dword v2, v[2:3] glc
; SI-NEXT: s_waitcnt vmcnt(0)
+; SI-NEXT: v_mov_b32_e32 v5, s7
; SI-NEXT: v_add_i32_e32 v0, vcc, s6, v4
; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v5, vcc
; SI-NEXT: flat_load_dword v0, v[0:1] glc
@@ -7235,11 +7235,11 @@ define amdgpu_kernel void @v_fneg_inlineasm_multi_use_src_f32(ptr addrspace(1) %
; VI-NEXT: v_mov_b32_e32 v3, s5
; VI-NEXT: v_add_u32_e32 v2, vcc, s4, v4
; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc
-; VI-NEXT: v_mov_b32_e32 v5, s7
; VI-NEXT: flat_load_dword v6, v[0:1] glc
; VI-NEXT: s_waitcnt vmcnt(0)
; VI-NEXT: flat_load_dword v2, v[2:3] glc
; VI-NEXT: s_waitcnt vmcnt(0)
+; VI-NEXT: v_mov_b32_e32 v5, s7
; VI-NEXT: v_add_u32_e32 v0, vcc, s6, v4
; VI-NEXT: v_addc_u32_e32 v1, vcc, 0, v5, vcc
; VI-NEXT: flat_load_dword v0, v[0:1] glc
diff --git a/llvm/test/CodeGen/AMDGPU/gds-allocation.ll b/llvm/test/CodeGen/AMDGPU/gds-allocation.ll
index 5c92976734c8be..ca1567b1e724d9 100644
--- a/llvm/test/CodeGen/AMDGPU/gds-allocation.ll
+++ b/llvm/test/CodeGen/AMDGPU/gds-allocation.ll
@@ -107,9 +107,9 @@ define amdgpu_kernel void @gds_extern_align(ptr addrspace(1) %out, ptr addrspace
; GCN-LABEL: gds_extern_align:
; GCN: ; %bb.0:
; GCN-NEXT: s_load_dword s0, s[4:5], 0x8
+; GCN-NEXT: s_movk_i32 s1, 0x400
; GCN-NEXT: v_mov_b32_e32 v0, 5
; GCN-NEXT: s_movk_i32 m0, 0x401
-; GCN-NEXT: s_movk_i32 s1, 0x400
; GCN-NEXT: ;;#ASMSTART
; GCN-NEXT: ; use s1
; GCN-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll b/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
index 7d028b383818de..ad0ac3c6467da2 100644
--- a/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
+++ b/llvm/test/CodeGen/AMDGPU/gfx-callable-preserved-registers.ll
@@ -258,12 +258,12 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_s31(ptr addrspace(1)
; GFX10-NEXT: s_mov_b32 s4, s31
; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
; GFX10-NEXT: s_mov_b32 s31, s4
+; GFX10-NEXT: v_readlane_b32 s4, v40, 0
; GFX10-NEXT: ;;#ASMSTART
; GFX10-NEXT: ; use s31
; GFX10-NEXT: ;;#ASMEND
; GFX10-NEXT: v_readlane_b32 s30, v40, 1
; GFX10-NEXT: v_readlane_b32 s31, v40, 2
-; GFX10-NEXT: v_readlane_b32 s4, v40, 0
; GFX10-NEXT: s_mov_b32 s32, s33
; GFX10-NEXT: v_readlane_b32 s34, v40, 3
; GFX10-NEXT: s_or_saveexec_b32 s35, -1
@@ -295,12 +295,12 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_s31(ptr addrspace(1)
; GFX11-NEXT: s_mov_b32 s4, s31
; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
; GFX11-NEXT: s_mov_b32 s31, s4
+; GFX11-NEXT: v_readlane_b32 s4, v40, 0
; GFX11-NEXT: ;;#ASMSTART
; GFX11-NEXT: ; use s31
; GFX11-NEXT: ;;#ASMEND
; GFX11-NEXT: v_readlane_b32 s30, v40, 1
; GFX11-NEXT: v_readlane_b32 s31, v40, 2
-; GFX11-NEXT: v_readlane_b32 s4, v40, 0
; GFX11-NEXT: s_mov_b32 s32, s33
; GFX11-NEXT: v_readlane_b32 s0, v40, 3
; GFX11-NEXT: s_or_saveexec_b32 s1, -1
@@ -374,14 +374,14 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_v31(ptr addrspace(1)
; GFX10-NEXT: v_mov_b32_e32 v40, v31
; GFX10-NEXT: s_swappc_b64 s[30:31], s[34:35]
; GFX10-NEXT: v_mov_b32_e32 v31, v40
-; GFX10-NEXT: ;;#ASMSTART
-; GFX10-NEXT: ; use v31
-; GFX10-NEXT: ;;#ASMEND
-; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GFX10-NEXT: v_readlane_b32 s30, v41, 0
; GFX10-NEXT: v_readlane_b32 s31, v41, 1
; GFX10-NEXT: s_mov_b32 s32, s33
; GFX10-NEXT: v_readlane_b32 s34, v41, 2
+; GFX10-NEXT: ;;#ASMSTART
+; GFX10-NEXT: ; use v31
+; GFX10-NEXT: ;;#ASMEND
+; GFX10-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload
; GFX10-NEXT: s_or_saveexec_b32 s35, -1
; GFX10-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
; GFX10-NEXT: s_waitcnt_depctr depctr_vm_vsrc(0)
@@ -412,14 +412,14 @@ define amdgpu_gfx void @test_call_void_func_void_mayclobber_v31(ptr addrspace(1)
; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX11-NEXT: v_mov_b32_e32 v31, v40
-; GFX11-NEXT: ;;#ASMSTART
-; GFX11-NEXT: ; use v31
-; GFX11-NEXT: ;;#ASMEND
-; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
; GFX11-NEXT: v_readlane_b32 s30, v41, 0
; GFX11-NEXT: v_readlane_b32 s31, v41, 1
; GFX11-NEXT: s_mov_b32 s32, s33
; GFX11-NEXT: v_readlane_b32 s0, v41, 2
+; GFX11-NEXT: ;;#ASMSTART
+; GFX11-NEXT: ; use v31
+; GFX11-NEXT: ;;#ASMEND
+; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload
; GFX11-NEXT: s_or_saveexec_b32 s1, -1
; GFX11-NEXT: scratch_load_b32 v41, off, s33 offset:4 ; 4-byte Folded Reload
; GFX11-NEXT: s_mov_b32 exec_lo, s1
@@ -644,8 +644,8 @@ define amdgpu_gfx void @test_call_void_func_void_preserves_s34(ptr addrspace(1)
; GFX11-NEXT: ;;#ASMEND
; GFX11-NEXT: s_mov_b32 s4, s34
; GFX11-NEXT: s_swappc_b64 s[30:31], s[0:1]
-; GFX11-NEXT: v_readlane_b32 s30, v40, 1
; GFX11-NEXT: s_mov_b32 s34, s4
+; GFX11-NEXT: v_readlane_b32 s30, v40, 1
; GFX11-NEXT: ;;#ASMSTART
; GFX11-NEXT: ; use s34
; GFX11-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll b/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll
index 2823adab8d63b9..4ac9cd9a907674 100644
--- a/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll
+++ b/llvm/test/CodeGen/AMDGPU/inline-asm-vgpr-sgpr-copy.ll
@@ -10,11 +10,11 @@ define <4 x float> @test_sgpr_constraint_bug(ptr addrspace(5) %buf_desc_ptr) {
; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; CHECK-NEXT: scratch_load_dwordx4 v[4:7], v0, off
; CHECK-NEXT: v_mov_b32_e32 v0, 0
+; CHECK-NEXT: s_mov_b64 s[4:5], exec
+; CHECK-NEXT: v_mov_b32_e32 v8, 1
; CHECK-NEXT: v_mov_b32_e32 v1, v0
; CHECK-NEXT: v_mov_b32_e32 v2, v0
; CHECK-NEXT: v_mov_b32_e32 v3, v0
-; CHECK-NEXT: s_mov_b64 s[4:5], exec
-; CHECK-NEXT: v_mov_b32_e32 v8, 1
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: v_readfirstlane_b32 s3, v7
; CHECK-NEXT: v_readfirstlane_b32 s2, v6
diff --git a/llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir b/llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir
new file mode 100644
index 00000000000000..39dfe2f30554a9
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/inlineasm-ds-return-latency.mir
@@ -0,0 +1,47 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -run-pass=postmisched -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Explicit inline-asm inputs retain native DS return latency, including LDS
+# atomics, configurable GDS operations, and fixed-GDS results.
+
+---
+name: atomic_lds_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: atomic_lds_result:%bb.0
+ ; CHECK: SU(0): {{.*}}DS_ADD_RTN_U32_gfx9
+ ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $vgpr1, $exec
+ $vgpr2 = DS_ADD_RTN_U32_gfx9 $vgpr0, $vgpr1, 0, 0, implicit $exec :: (load store (s32), addrspace 3)
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
+
+---
+name: atomic_configurable_gds_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: atomic_configurable_gds_result:%bb.0
+ ; CHECK: SU(0): {{.*}}DS_ADD_RTN_U32
+ ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $vgpr1, $m0, $exec
+ $vgpr2 = DS_ADD_RTN_U32 $vgpr0, $vgpr1, 0, 1, implicit $m0, implicit $exec :: (load store (s32), addrspace 2)
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
+
+---
+name: fixed_gds_ordered_count_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: fixed_gds_ordered_count_result:%bb.0
+ ; CHECK: SU(0): {{.*}}DS_ORDERED_COUNT
+ ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $m0, $exec
+ $vgpr2 = DS_ORDERED_COUNT $vgpr0, 772, implicit $m0, implicit $exec :: (load store (s32), addrspace 2)
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir b/llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir
new file mode 100644
index 00000000000000..aa12bbd9d76beb
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/inlineasm-producer-latency.mir
@@ -0,0 +1,47 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Explicit inline-asm inputs retain modeled producer latency for non-DS
+# instructions, including arithmetic, global loads, and matrix instructions.
+
+---
+name: valu_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: valu_result:%bb.0
+ ; CHECK: SU(0): {{.*}}V_ADD_U32_e32
+ ; CHECK: SU(1): Data Latency=1 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $exec
+ $vgpr2 = V_ADD_U32_e32 1, $vgpr0, implicit $exec
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
+
+---
+name: vmem_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: vmem_result:%bb.0
+ ; CHECK: SU(0): {{.*}}GLOBAL_LOAD_DWORD
+ ; CHECK: SU(1): Data Latency=80 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0_vgpr1, $exec
+ $vgpr2 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec :: (load (s32), addrspace 1)
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
+
+---
+name: mfma_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: mfma_result:%bb.0
+ ; CHECK: SU(0): {{.*}}V_MFMA_F32_16X16X16F16_vgprcd_e64
+ ; CHECK: SU(1): Data Latency=4 Reg=$vgpr4_vgpr5_vgpr6_vgpr7
+ bb.0:
+ liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr8_vgpr9_vgpr10_vgpr11, $mode, $exec
+ $vgpr4_vgpr5_vgpr6_vgpr7 = V_MFMA_F32_16X16X16F16_vgprcd_e64 $vgpr0_vgpr1, $vgpr2_vgpr3, $vgpr8_vgpr9_vgpr10_vgpr11, 0, 0, 0, implicit $mode, implicit $exec
+ INLINEASM &"", attdialect, regdef:VReg_128_Align2, def $vgpr4_vgpr5_vgpr6_vgpr7, reguse tiedto:$0, $vgpr4_vgpr5_vgpr6_vgpr7(tied-def 3)
+ S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
index 0d94fa29aa229e..2268d95bd70839 100644
--- a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2i16.ll
@@ -155,8 +155,8 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_multi_use_hi_reg(ptr addrspac
; VI-NEXT: v_mov_b32_e32 v1, s1
; VI-NEXT: s_waitcnt lgkmcnt(0)
; VI-NEXT: v_mov_b32_e32 v3, s2
-; VI-NEXT: v_perm_b32 v2, s4, v3, v2
; VI-NEXT: s_lshr_b32 s0, s2, 16
+; VI-NEXT: v_perm_b32 v2, s4, v3, v2
; VI-NEXT: flat_store_dword v[0:1], v2
; VI-NEXT: ;;#ASMSTART
; VI-NEXT: ; use s0
@@ -178,8 +178,8 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_multi_use_hi_reg(ptr addrspac
; CI-NEXT: s_waitcnt lgkmcnt(0)
; CI-NEXT: s_and_b32 s1, s2, 0xffff0000
; CI-NEXT: s_or_b32 s0, s0, s1
-; CI-NEXT: v_mov_b32_e32 v2, s0
; CI-NEXT: s_lshr_b32 s2, s2, 16
+; CI-NEXT: v_mov_b32_e32 v2, s0
; CI-NEXT: flat_store_dword v[0:1], v2
; CI-NEXT: ;;#ASMSTART
; CI-NEXT: ; use s2
@@ -409,8 +409,8 @@ define amdgpu_kernel void @s_insertelement_v2i16_0_reghi_both_multi_use_1(ptr ad
; VI-NEXT: s_lshr_b32 s0, s4, 16
; VI-NEXT: s_waitcnt lgkmcnt(0)
; VI-NEXT: v_mov_b32_e32 v3, s2
-; VI-NEXT: v_perm_b32 v2, s4, v3, v2
; VI-NEXT: s_lshr_b32 s1, s2, 16
+; VI-NEXT: v_perm_b32 v2, s4, v3, v2
; VI-NEXT: flat_store_dword v[0:1], v2
; VI-NEXT: ;;#ASMSTART
; VI-NEXT: ; use s0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
index 8dfb872b1fa323..3be6730ff5a211 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
@@ -19,11 +19,11 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
; SI-SDAG-NEXT: v_mov_b32_e32 v0, s2
; SI-SDAG-NEXT: v_mov_b32_e32 v1, s3
; SI-SDAG-NEXT: s_movk_i32 s0, 0x64
-; SI-SDAG-NEXT: s_mov_b32 s5, s1
; SI-SDAG-NEXT: ;;#ASMSTART
; SI-SDAG-NEXT: ; def v[4:5], use v[0:1]
; SI-SDAG-NEXT: ;;#ASMEND
; SI-SDAG-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[4:5], s0, v[2:3]
+; SI-SDAG-NEXT: s_mov_b32 s5, s1
; SI-SDAG-NEXT: v_mov_b32_e32 v5, v1
; SI-SDAG-NEXT: v_mov_b32_e32 v4, v0
; SI-SDAG-NEXT: ;;#ASMSTART
@@ -46,9 +46,9 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
; SI-GISEL-NEXT: ;;#ASMEND
; SI-GISEL-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[4:5], v6, v[2:3]
; SI-GISEL-NEXT: s_mov_b32 s2, -1
-; SI-GISEL-NEXT: s_mov_b32 s3, 0xf000
; SI-GISEL-NEXT: v_mov_b32_e32 v5, v1
; SI-GISEL-NEXT: v_mov_b32_e32 v4, v0
+; SI-GISEL-NEXT: s_mov_b32 s3, 0xf000
; SI-GISEL-NEXT: ;;#ASMSTART
; SI-GISEL-NEXT: ;; force constraint
; SI-GISEL-NEXT: ;;#ASMEND
@@ -64,12 +64,12 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
; VI-SDAG-NEXT: s_waitcnt lgkmcnt(0)
; VI-SDAG-NEXT: v_mov_b32_e32 v0, s2
; VI-SDAG-NEXT: v_mov_b32_e32 v1, s3
-; VI-SDAG-NEXT: v_mov_b32_e32 v2, s0
-; VI-SDAG-NEXT: v_mov_b32_e32 v3, s1
; VI-SDAG-NEXT: ;;#ASMSTART
; VI-SDAG-NEXT: ; def v[4:5], use v[0:1]
; VI-SDAG-NEXT: ;;#ASMEND
; VI-SDAG-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[4:5], s4, v[6:7]
+; VI-SDAG-NEXT: v_mov_b32_e32 v2, s0
+; VI-SDAG-NEXT: v_mov_b32_e32 v3, s1
; VI-SDAG-NEXT: v_mov_b32_e32 v5, v1
; VI-SDAG-NEXT: v_mov_b32_e32 v4, v0
; VI-SDAG-NEXT: ;;#ASMSTART
@@ -91,10 +91,10 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8(ptr addrspace(1) %out, i64 %src) {
; VI-GISEL-NEXT: ; def v[4:5], use v[0:1]
; VI-GISEL-NEXT: ;;#ASMEND
; VI-GISEL-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[4:5], v6, v[2:3]
-; VI-GISEL-NEXT: v_mov_b32_e32 v3, s1
-; VI-GISEL-NEXT: v_mov_b32_e32 v2, s0
; VI-GISEL-NEXT: v_mov_b32_e32 v5, v1
; VI-GISEL-NEXT: v_mov_b32_e32 v4, v0
+; VI-GISEL-NEXT: v_mov_b32_e32 v3, s1
+; VI-GISEL-NEXT: v_mov_b32_e32 v2, s0
; VI-GISEL-NEXT: ;;#ASMSTART
; VI-GISEL-NEXT: ;; force constraint
; VI-GISEL-NEXT: ;;#ASMEND
@@ -127,12 +127,12 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8_non_immediate(ptr addrspace(1) %out
; SI-SDAG-NEXT: ;;#ASMEND
; SI-SDAG-NEXT: v_mov_b32_e32 v0, s8
; SI-SDAG-NEXT: v_mov_b32_e32 v1, s9
-; SI-SDAG-NEXT: s_mov_b32 s4, s0
-; SI-SDAG-NEXT: s_mov_b32 s5, s1
; SI-SDAG-NEXT: ;;#ASMSTART
; SI-SDAG-NEXT: ; def v[6:7], use v[0:1]
; SI-SDAG-NEXT: ;;#ASMEND
; SI-SDAG-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[2:3], v4, v[6:7]
+; SI-SDAG-NEXT: s_mov_b32 s4, s0
+; SI-SDAG-NEXT: s_mov_b32 s5, s1
; SI-SDAG-NEXT: v_mov_b32_e32 v3, v1
; SI-SDAG-NEXT: v_mov_b32_e32 v2, v0
; SI-SDAG-NEXT: ;;#ASMSTART
@@ -162,10 +162,10 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8_non_immediate(ptr addrspace(1) %out
; SI-GISEL-NEXT: ; def v[6:7], use v[0:1]
; SI-GISEL-NEXT: ;;#ASMEND
; SI-GISEL-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[2:3], v4, v[6:7]
-; SI-GISEL-NEXT: s_mov_b32 s2, -1
-; SI-GISEL-NEXT: s_mov_b32 s3, 0xf000
; SI-GISEL-NEXT: v_mov_b32_e32 v3, v1
; SI-GISEL-NEXT: v_mov_b32_e32 v2, v0
+; SI-GISEL-NEXT: s_mov_b32 s2, -1
+; SI-GISEL-NEXT: s_mov_b32 s3, 0xf000
; SI-GISEL-NEXT: ;;#ASMSTART
; SI-GISEL-NEXT: ;; force constraint
; SI-GISEL-NEXT: ;;#ASMEND
@@ -189,12 +189,12 @@ define amdgpu_kernel void @v_mqsad_pk_u16_u8_non_immediate(ptr addrspace(1) %out
; VI-SDAG-NEXT: ;;#ASMEND
; VI-SDAG-NEXT: v_mov_b32_e32 v0, s4
; VI-SDAG-NEXT: v_mov_b32_e32 v1, s5
-; VI-SDAG-NEXT: v_mov_b32_e32 v8, s0
-; VI-SDAG-NEXT: v_mov_b32_e32 v9, s1
; VI-SDAG-NEXT: ;;#ASMSTART
; VI-SDAG-NEXT: ; def v[6:7], use v[0:1]
; VI-SDAG-NEXT: ;;#ASMEND
; VI-SDAG-NEXT: v_mqsad_pk_u16_u8 v[0:1], v[2:3], v4, v[6:7]
+; VI-SDAG-NEXT: v_mov_b32_e32 v8, s0
+; VI-SDAG-NEXT: v_mov_b32_e32 v9, s1
; VI-SDAG-NEXT: v_mov_b32_e32 v3, v1
; VI-SDAG-NEXT: v_mov_b32_e32 v2, v0
; VI-SDAG-NEXT: ;;#ASMSTART
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
index ec3b61a81f2e80..4c027cd21278c0 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.u32.u8.ll
@@ -42,13 +42,13 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_integer_immediate(ptr addrspace
; GFX803-GISEL-NEXT: s_mov_b32 s3, 40
; GFX803-GISEL-NEXT: s_mov_b32 s2, 30
; GFX803-GISEL-NEXT: s_mov_b32 s1, 20
-; GFX803-GISEL-NEXT: s_mov_b32 s0, 10
; GFX803-GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GFX803-GISEL-NEXT: v_mov_b32_e32 v0, s10
; GFX803-GISEL-NEXT: v_mov_b32_e32 v1, s11
; GFX803-GISEL-NEXT: ;;#ASMSTART
; GFX803-GISEL-NEXT: ; def v[2:3], use v[0:1]
; GFX803-GISEL-NEXT: ;;#ASMEND
+; GFX803-GISEL-NEXT: s_mov_b32 s0, 10
; GFX803-GISEL-NEXT: v_mov_b32_e32 v0, v2
; GFX803-GISEL-NEXT: v_mov_b32_e32 v1, v3
; GFX803-GISEL-NEXT: v_mov_b32_e32 v2, s4
@@ -268,13 +268,13 @@ define amdgpu_kernel void @v_mqsad_u32_u8_inline_fp_immediate(ptr addrspace(1) %
; GFX803-GISEL-NEXT: s_mov_b32 s1, 0
; GFX803-GISEL-NEXT: s_mov_b32 s0, 1.0
; GFX803-GISEL-NEXT: s_mov_b32 s2, s1
-; GFX803-GISEL-NEXT: s_mov_b32 s3, s1
; GFX803-GISEL-NEXT: s_waitcnt lgkmcnt(0)
; GFX803-GISEL-NEXT: v_mov_b32_e32 v0, s10
; GFX803-GISEL-NEXT: v_mov_b32_e32 v1, s11
; GFX803-GISEL-NEXT: ;;#ASMSTART
; GFX803-GISEL-NEXT: ; def v[2:3], use v[0:1]
; GFX803-GISEL-NEXT: ;;#ASMEND
+; GFX803-GISEL-NEXT: s_mov_b32 s3, s1
; GFX803-GISEL-NEXT: v_mov_b32_e32 v0, v2
; GFX803-GISEL-NEXT: v_mov_b32_e32 v1, v3
; GFX803-GISEL-NEXT: v_mov_b32_e32 v2, s4
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
index fd341003bce63d..3672ea6b063b36 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
@@ -667,8 +667,8 @@ define amdgpu_kernel void @test_readfirstlane_fi(ptr addrspace(1) %out) #0 {
; CHECK-SDAG-LABEL: test_readfirstlane_fi:
; CHECK-SDAG: ; %bb.0:
; CHECK-SDAG-NEXT: s_add_u32 s0, s0, s17
-; CHECK-SDAG-NEXT: s_addc_u32 s1, s1, 0
; CHECK-SDAG-NEXT: s_mov_b32 s4, 0
+; CHECK-SDAG-NEXT: s_addc_u32 s1, s1, 0
; CHECK-SDAG-NEXT: ;;#ASMSTART
; CHECK-SDAG-NEXT: ; use s4
; CHECK-SDAG-NEXT: ;;#ASMEND
@@ -677,8 +677,8 @@ define amdgpu_kernel void @test_readfirstlane_fi(ptr addrspace(1) %out) #0 {
; CHECK-GISEL-LABEL: test_readfirstlane_fi:
; CHECK-GISEL: ; %bb.0:
; CHECK-GISEL-NEXT: s_add_u32 s0, s0, s17
-; CHECK-GISEL-NEXT: s_addc_u32 s1, s1, 0
; CHECK-GISEL-NEXT: s_mov_b32 s4, 0
+; CHECK-GISEL-NEXT: s_addc_u32 s1, s1, 0
; CHECK-GISEL-NEXT: ;;#ASMSTART
; CHECK-GISEL-NEXT: ; use s4
; CHECK-GISEL-NEXT: ;;#ASMEND
@@ -1005,6 +1005,13 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
; CHECK-SDAG-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:4
; CHECK-SDAG-NEXT: buffer_load_dword v27, off, s[0:3], s32
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s64, v30
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s63, v29
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s62, v28
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s60, v26
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s59, v25
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s58, v24
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s57, v23
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s56, v22
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s55, v21
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s54, v20
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s53, v19
@@ -1013,17 +1020,6 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s50, v16
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s49, v15
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s48, v14
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s39, v5
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s38, v4
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s37, v3
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s36, v2
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s63, v29
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s62, v28
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s60, v26
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s59, v25
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s58, v24
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s57, v23
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s56, v22
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s47, v13
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s46, v12
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s45, v11
@@ -1032,6 +1028,10 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s42, v8
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s41, v7
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s40, v6
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s39, v5
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s38, v4
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s37, v3
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s36, v2
; CHECK-SDAG-NEXT: s_waitcnt vmcnt(2)
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s67, v0
; CHECK-SDAG-NEXT: s_waitcnt vmcnt(1)
@@ -1092,15 +1092,6 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s37, v3
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s38, v4
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s39, v5
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s48, v14
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s49, v15
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s50, v16
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s51, v17
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s52, v18
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s53, v19
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s54, v20
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s55, v21
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s64, v30
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s40, v6
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s41, v7
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s42, v8
@@ -1109,6 +1100,14 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s45, v11
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s46, v12
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s47, v13
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s48, v14
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s49, v15
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s50, v16
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s51, v17
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s52, v18
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s53, v19
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s54, v20
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s55, v21
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s56, v22
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s57, v23
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s58, v24
@@ -1117,6 +1116,7 @@ define void @test_readfirstlane_v32f32(ptr addrspace(1) %out, <32 x float> %src)
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s61, v27
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s62, v28
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s63, v29
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s64, v30
; CHECK-GISEL-NEXT: s_waitcnt vmcnt(2)
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s65, v0
; CHECK-GISEL-NEXT: s_waitcnt vmcnt(1)
@@ -1450,6 +1450,13 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
; CHECK-SDAG-NEXT: buffer_load_dword v1, off, s[0:3], s32 offset:4
; CHECK-SDAG-NEXT: buffer_load_dword v27, off, s[0:3], s32
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s64, v30
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s63, v29
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s62, v28
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s60, v26
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s59, v25
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s58, v24
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s57, v23
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s56, v22
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s55, v21
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s54, v20
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s53, v19
@@ -1458,17 +1465,6 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s50, v16
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s49, v15
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s48, v14
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s39, v5
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s38, v4
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s37, v3
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s36, v2
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s63, v29
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s62, v28
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s60, v26
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s59, v25
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s58, v24
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s57, v23
-; CHECK-SDAG-NEXT: v_readfirstlane_b32 s56, v22
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s47, v13
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s46, v12
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s45, v11
@@ -1477,6 +1473,10 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s42, v8
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s41, v7
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s40, v6
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s39, v5
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s38, v4
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s37, v3
+; CHECK-SDAG-NEXT: v_readfirstlane_b32 s36, v2
; CHECK-SDAG-NEXT: s_waitcnt vmcnt(2)
; CHECK-SDAG-NEXT: v_readfirstlane_b32 s67, v0
; CHECK-SDAG-NEXT: s_waitcnt vmcnt(1)
@@ -1537,15 +1537,6 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s37, v3
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s38, v4
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s39, v5
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s48, v14
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s49, v15
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s50, v16
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s51, v17
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s52, v18
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s53, v19
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s54, v20
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s55, v21
-; CHECK-GISEL-NEXT: v_readfirstlane_b32 s64, v30
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s40, v6
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s41, v7
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s42, v8
@@ -1554,6 +1545,14 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s45, v11
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s46, v12
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s47, v13
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s48, v14
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s49, v15
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s50, v16
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s51, v17
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s52, v18
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s53, v19
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s54, v20
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s55, v21
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s56, v22
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s57, v23
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s58, v24
@@ -1562,6 +1561,7 @@ define void @test_readfirstlane_v32i32(ptr addrspace(1) %out, <32 x i32> %src) #
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s61, v27
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s62, v28
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s63, v29
+; CHECK-GISEL-NEXT: v_readfirstlane_b32 s64, v30
; CHECK-GISEL-NEXT: s_waitcnt vmcnt(2)
; CHECK-GISEL-NEXT: v_readfirstlane_b32 s65, v0
; CHECK-GISEL-NEXT: s_waitcnt vmcnt(1)
diff --git a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll
index 32a288e770416d..068c2eacd6ed8e 100644
--- a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll
+++ b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.gfx10.ll
@@ -148,8 +148,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc() #0 {
; GFX8-NEXT: v_lshrrev_b32_e64 v0, 6, s32
; GFX8-NEXT: s_movk_i32 vcc_lo, 0x4040
; GFX8-NEXT: v_add_u32_e32 v0, vcc, vcc_lo, v0
-; GFX8-NEXT: v_readfirstlane_b32 s55, v0
; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX8-NEXT: v_readfirstlane_b32 s55, v0
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use s55, scc
; GFX8-NEXT: ;;#ASMEND
@@ -176,8 +176,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: v_lshrrev_b32_e64 v0, 6, s32
; GFX900-NEXT: v_add_u32_e32 v0, 0x4040, v0
-; GFX900-NEXT: v_readfirstlane_b32 s55, v0
; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX900-NEXT: v_readfirstlane_b32 s55, v0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s55, scc
; GFX900-NEXT: ;;#ASMEND
@@ -320,13 +320,13 @@ define void @scalar_mov_materializes_frame_index_dead_scc() #0 {
; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
; GFX12-NEXT: s_mov_b32 exec_lo, s0
; GFX12-NEXT: v_writelane_b32 v1, s55, 0
-; GFX12-NEXT: s_add_co_i32 s0, s32, 0x4000
; GFX12-NEXT: v_mov_b32_e32 v0, s32
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_mov_b32 s55, s0
+; GFX12-NEXT: s_add_co_i32 s0, s32, 0x4000
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use alloca0 v0
; GFX12-NEXT: ;;#ASMEND
+; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
+; GFX12-NEXT: s_mov_b32 s55, s0
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use s55
; GFX12-NEXT: ;;#ASMEND
@@ -346,10 +346,10 @@ define void @scalar_mov_materializes_frame_index_dead_scc() #0 {
; GFX8-NEXT: buffer_store_dword v1, off, s[0:3], s6 ; 4-byte Folded Spill
; GFX8-NEXT: s_mov_b64 exec, s[4:5]
; GFX8-NEXT: v_writelane_b32 v1, s55, 0
-; GFX8-NEXT: s_lshr_b32 s55, s32, 6
; GFX8-NEXT: v_lshrrev_b32_e64 v0, 6, s32
-; GFX8-NEXT: s_addk_i32 s55, 0x4040
+; GFX8-NEXT: s_lshr_b32 s55, s32, 6
; GFX8-NEXT: v_add_u32_e32 v0, vcc, 64, v0
+; GFX8-NEXT: s_addk_i32 s55, 0x4040
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use alloca0 v0
; GFX8-NEXT: ;;#ASMEND
@@ -372,10 +372,10 @@ define void @scalar_mov_materializes_frame_index_dead_scc() #0 {
; GFX900-NEXT: buffer_store_dword v1, off, s[0:3], s6 ; 4-byte Folded Spill
; GFX900-NEXT: s_mov_b64 exec, s[4:5]
; GFX900-NEXT: v_writelane_b32 v1, s55, 0
-; GFX900-NEXT: s_lshr_b32 s55, s32, 6
; GFX900-NEXT: v_lshrrev_b32_e64 v0, 6, s32
-; GFX900-NEXT: s_addk_i32 s55, 0x4040
+; GFX900-NEXT: s_lshr_b32 s55, s32, 6
; GFX900-NEXT: v_add_u32_e32 v0, 64, v0
+; GFX900-NEXT: s_addk_i32 s55, 0x4040
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use alloca0 v0
; GFX900-NEXT: ;;#ASMEND
@@ -585,8 +585,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc_fp() #1 {
; GFX8-NEXT: v_lshrrev_b32_e64 v0, 6, s33
; GFX8-NEXT: s_movk_i32 vcc_lo, 0x4040
; GFX8-NEXT: v_add_u32_e32 v0, vcc, vcc_lo, v0
-; GFX8-NEXT: v_readfirstlane_b32 s55, v0
; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX8-NEXT: v_readfirstlane_b32 s55, v0
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use s55, scc
; GFX8-NEXT: ;;#ASMEND
@@ -618,8 +618,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc_fp() #1 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: v_lshrrev_b32_e64 v0, 6, s33
; GFX900-NEXT: v_add_u32_e32 v0, 0x4040, v0
-; GFX900-NEXT: v_readfirstlane_b32 s55, v0
; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX900-NEXT: v_readfirstlane_b32 s55, v0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s55, scc
; GFX900-NEXT: ;;#ASMEND
@@ -784,8 +784,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc_small_offset()
; GFX8-NEXT: v_lshrrev_b32_e64 v1, 6, s32
; GFX8-NEXT: s_mov_b32 vcc_lo, 64
; GFX8-NEXT: v_add_u32_e32 v1, vcc, vcc_lo, v1
-; GFX8-NEXT: v_readfirstlane_b32 s55, v1
; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX8-NEXT: v_readfirstlane_b32 s55, v1
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use s55, scc
; GFX8-NEXT: ;;#ASMEND
@@ -807,8 +807,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc_small_offset()
; GFX900-NEXT: v_writelane_b32 v0, s55, 0
; GFX900-NEXT: v_lshrrev_b32_e64 v1, 6, s32
; GFX900-NEXT: v_add_u32_e32 v1, 64, v1
-; GFX900-NEXT: v_readfirstlane_b32 s55, v1
; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX900-NEXT: v_readfirstlane_b32 s55, v1
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s55, scc
; GFX900-NEXT: ;;#ASMEND
@@ -1143,8 +1143,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc_small_offset_fp
; GFX8-NEXT: v_lshrrev_b32_e64 v1, 6, s33
; GFX8-NEXT: s_mov_b32 vcc_lo, 64
; GFX8-NEXT: v_add_u32_e32 v1, vcc, vcc_lo, v1
-; GFX8-NEXT: v_readfirstlane_b32 s55, v1
; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX8-NEXT: v_readfirstlane_b32 s55, v1
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use s55, scc
; GFX8-NEXT: ;;#ASMEND
@@ -1171,8 +1171,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc_small_offset_fp
; GFX900-NEXT: v_writelane_b32 v0, s55, 0
; GFX900-NEXT: v_lshrrev_b32_e64 v1, 6, s33
; GFX900-NEXT: v_add_u32_e32 v1, 64, v1
-; GFX900-NEXT: v_readfirstlane_b32 s55, v1
; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
+; GFX900-NEXT: v_readfirstlane_b32 s55, v1
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s55, scc
; GFX900-NEXT: ;;#ASMEND
@@ -1316,12 +1316,12 @@ define void @scalar_mov_materializes_frame_index_available_scc_small_offset_fp()
; GFX12-NEXT: s_addk_co_i32 s32, 0x4040
; GFX12-NEXT: v_writelane_b32 v0, s55, 0
; GFX12-NEXT: s_mov_b32 s55, s33
+; GFX12-NEXT: s_mov_b32 s32, s33
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use s55
; GFX12-NEXT: ;;#ASMEND
; GFX12-NEXT: s_delay_alu instid0(VALU_DEP_1)
; GFX12-NEXT: v_readlane_b32 s55, v0, 0
-; GFX12-NEXT: s_mov_b32 s32, s33
; GFX12-NEXT: s_xor_saveexec_b32 s1, -1
; GFX12-NEXT: scratch_load_b32 v0, off, s33 offset:16384 ; 4-byte Folded Reload
; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
@@ -1508,8 +1508,8 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
; GFX12-NEXT: s_mov_b32 exec_lo, s0
; GFX12-NEXT: v_writelane_b32 v1, s55, 0
-; GFX12-NEXT: s_add_co_i32 s55, s32, 0x43ec
; GFX12-NEXT: v_mov_b32_e32 v0, s32
+; GFX12-NEXT: s_add_co_i32 s55, s32, 0x43ec
; GFX12-NEXT: s_and_b32 s0, 0, exec_lo
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use alloca0 v0
@@ -1537,10 +1537,10 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
; GFX8-NEXT: s_add_i32 s55, s4, 0x442c
; GFX8-NEXT: v_lshrrev_b32_e64 v0, 6, s32
; GFX8-NEXT: v_add_u32_e32 v0, vcc, 64, v0
+; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use alloca0 v0
; GFX8-NEXT: ;;#ASMEND
-; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use s55, scc
; GFX8-NEXT: ;;#ASMEND
@@ -1564,10 +1564,10 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
; GFX900-NEXT: s_add_i32 s55, s4, 0x442c
; GFX900-NEXT: v_lshrrev_b32_e64 v0, 6, s32
; GFX900-NEXT: v_add_u32_e32 v0, 64, v0
+; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use alloca0 v0
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s55, scc
; GFX900-NEXT: ;;#ASMEND
@@ -1590,10 +1590,10 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_immoffset(
; GFX942-NEXT: s_add_i32 s55, s32, 0x442c
; GFX942-NEXT: s_add_i32 s0, s32, 64
; GFX942-NEXT: v_mov_b32_e32 v0, s0
+; GFX942-NEXT: s_and_b64 s[0:1], 0, exec
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use alloca0 v0
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: s_and_b64 s[0:1], 0, exec
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s55, scc
; GFX942-NEXT: ;;#ASMEND
@@ -1711,14 +1711,14 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
; GFX12-NEXT: s_mov_b32 exec_lo, s1
; GFX12-NEXT: v_writelane_b32 v1, s55, 0
-; GFX12-NEXT: s_add_co_i32 s1, s32, 0x4000
; GFX12-NEXT: v_mov_b32_e32 v0, s32
-; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
-; GFX12-NEXT: s_lshl2_add_u32 s55, s0, s1
-; GFX12-NEXT: s_and_b32 s0, 0, exec_lo
+; GFX12-NEXT: s_add_co_i32 s1, s32, 0x4000
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use alloca0 v0
; GFX12-NEXT: ;;#ASMEND
+; GFX12-NEXT: s_wait_alu depctr_sa_sdst(0)
+; GFX12-NEXT: s_lshl2_add_u32 s55, s0, s1
+; GFX12-NEXT: s_and_b32 s0, 0, exec_lo
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use s55, scc
; GFX12-NEXT: ;;#ASMEND
@@ -1744,10 +1744,10 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
; GFX8-NEXT: s_addk_i32 s55, 0x4040
; GFX8-NEXT: v_lshrrev_b32_e64 v0, 6, s32
; GFX8-NEXT: v_add_u32_e32 v0, vcc, 64, v0
+; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use alloca0 v0
; GFX8-NEXT: ;;#ASMEND
-; GFX8-NEXT: s_and_b64 s[4:5], 0, exec
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use s55, scc
; GFX8-NEXT: ;;#ASMEND
@@ -1772,10 +1772,10 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
; GFX900-NEXT: s_lshl2_add_u32 s55, s16, s4
; GFX900-NEXT: v_lshrrev_b32_e64 v0, 6, s32
; GFX900-NEXT: v_add_u32_e32 v0, 64, v0
+; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use alloca0 v0
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: s_and_b64 s[4:5], 0, exec
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s55, scc
; GFX900-NEXT: ;;#ASMEND
@@ -1799,10 +1799,10 @@ define void @scalar_mov_materializes_frame_index_unavailable_scc__gep_sgpr_offse
; GFX942-NEXT: s_lshl2_add_u32 s55, s0, s1
; GFX942-NEXT: s_add_i32 s0, s32, 64
; GFX942-NEXT: v_mov_b32_e32 v0, s0
+; GFX942-NEXT: s_and_b64 s[0:1], 0, exec
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use alloca0 v0
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: s_and_b64 s[0:1], 0, exec
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s55, scc
; GFX942-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll
index f451c192db14dd..c1dc22ad310442 100644
--- a/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll
+++ b/llvm/test/CodeGen/AMDGPU/materialize-frame-index-sgpr.ll
@@ -1043,8 +1043,8 @@ define void @scalar_mov_materializes_frame_index_no_live_scc_no_live_sgprs__lowe
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; def s[0:15], s[16:31], s[32:47], s[48:55], s[56:57], s58, v[0:15], v[16:20], vcc
; GFX12-NEXT: ;;#ASMEND
-; GFX12-NEXT: s_and_b32 s59, 0, exec_lo
; GFX12-NEXT: s_mov_b32 s54, s32
+; GFX12-NEXT: s_and_b32 s59, 0, exec_lo
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use s[0:15], s[16:31], s[32:47], s[48:55], s[56:57], s58, v[0:15], v[16:20], vcc, s54, scc
; GFX12-NEXT: ;;#ASMEND
@@ -1604,8 +1604,8 @@ define void @scalar_mov_materializes_frame_index_no_live_scc_no_live_sgprs_gep_i
; GFX12-NEXT: v_writelane_b32 v22, s55, 14
; GFX12-NEXT: v_writelane_b32 v22, s30, 15
; GFX12-NEXT: v_writelane_b32 v22, s31, 16
-; GFX12-NEXT: s_add_co_i32 s58, s32, 0x4200
; GFX12-NEXT: v_mov_b32_e32 v0, s32
+; GFX12-NEXT: s_add_co_i32 s58, s32, 0x4200
; GFX12-NEXT: s_and_b32 s0, 0, exec_lo
; GFX12-NEXT: ;;#ASMSTART
; GFX12-NEXT: ; use alloca0 v0
diff --git a/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir
new file mode 100644
index 00000000000000..170291fc80b33f
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-flags.mir
@@ -0,0 +1,48 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -verify-machineinstrs -verify-misched -debug-only=machine-scheduler %s -o /dev/null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Encoded DS reads and reads without MMOs retain the modeled result latency
+# at explicit inline-asm inputs. Generic side-effect/order flags do not change
+# the latency of a register result.
+
+---
+name: encoded_read
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: encoded_read:%bb.0
+ ; CHECK: SU(0): {{.*}}DS_READ_B32_vi_gfx9
+ ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $exec
+ $vgpr2 = DS_READ_B32_vi_gfx9 $vgpr0, 0, 0, implicit $exec :: (load (s32), addrspace 3)
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
+
+---
+name: missing_mmo
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: missing_mmo:%bb.0
+ ; CHECK: SU(0): {{.*}}DS_READ_B32_gfx9
+ ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $exec
+ $vgpr2 = DS_READ_B32_gfx9 $vgpr0, 0, 0, implicit $exec
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
+
+---
+name: atomic_lds_result
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: atomic_lds_result:%bb.0
+ ; CHECK: SU(0): {{.*}}DS_ADD_RTN_U32_gfx9
+ ; CHECK: SU(1): Data Latency=5 Reg=$vgpr2
+ bb.0:
+ liveins: $vgpr0, $vgpr1, $exec
+ $vgpr2 = DS_ADD_RTN_U32_gfx9 $vgpr0, $vgpr1, 0, 0, implicit $exec :: (load store (s32), addrspace 3)
+ INLINEASM &"v_mov_b32 $0, $1", attdialect, regdef:VGPR_32, def $vgpr12, reguse:VGPR_32, $vgpr2
+ S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir
new file mode 100644
index 00000000000000..92a6879290951b
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/postra-coexec-inlineasm-latency.mir
@@ -0,0 +1,93 @@
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -run-pass=postmisched -misched-cluster=false -verify-machineinstrs -verify-misched %s -o - | FileCheck %s
+
+# A real asm input must retain its native LDS producer's modeled latency in
+# default and coexec schedules. Check empty tied and nonempty inputs, including
+# IGLP.
+
+--- |
+ define amdgpu_kernel void @tied_input_default() { ret void }
+ define amdgpu_kernel void @tied_input_coexec() #0 { ret void }
+ define amdgpu_kernel void @explicit_input_coexec() #0 { ret void }
+ define amdgpu_kernel void @iglp_input_coexec() #0 { ret void }
+ attributes #0 = { "amdgpu-sched-strategy"="coexec" }
+...
+
+---
+name: tied_input_default
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: tied_input_default
+ ; CHECK: liveins: $vgpr0, $vgpr4, $exec
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $vgpr2_vgpr3 = DS_READ_B64_TR_B16
+ bb.0:
+ liveins: $vgpr0, $vgpr4, $exec
+
+ $vgpr2_vgpr3 = DS_READ_B64_TR_B16 $vgpr0, 64, 0, implicit $exec :: (load (s64), addrspace 3)
+ INLINEASM &"", attdialect, regdef:VReg_64_Align2, def $vgpr2_vgpr3, reguse tiedto:$0, $vgpr2_vgpr3(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ $vgpr5 = V_ADD_U32_e32 $vgpr2, $vgpr3, implicit $exec
+ S_ENDPGM 0
+...
+
+---
+name: tied_input_coexec
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: tied_input_coexec
+ ; CHECK: liveins: $vgpr0, $vgpr4, $exec
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $vgpr2_vgpr3 = DS_READ_B64_TR_B16
+ bb.0:
+ liveins: $vgpr0, $vgpr4, $exec
+
+ $vgpr2_vgpr3 = DS_READ_B64_TR_B16 $vgpr0, 64, 0, implicit $exec :: (load (s64), addrspace 3)
+ INLINEASM &"", attdialect, regdef:VReg_64_Align2, def $vgpr2_vgpr3, reguse tiedto:$0, $vgpr2_vgpr3(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ $vgpr5 = V_ADD_U32_e32 $vgpr2, $vgpr3, implicit $exec
+ S_ENDPGM 0
+...
+
+---
+name: explicit_input_coexec
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: explicit_input_coexec
+ ; CHECK: liveins: $vgpr0, $vgpr4, $exec
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: $vgpr2 = DS_READ_B32_gfx9
+ bb.0:
+ liveins: $vgpr0, $vgpr4, $exec
+
+ $vgpr2 = DS_READ_B32_gfx9 $vgpr0, 64, 0, implicit $exec :: (load (s32), addrspace 3)
+ INLINEASM &"v_add_u32 $0, $1, 1", attdialect, regdef:VGPR_32, def $vgpr6, reguse:VGPR_32, $vgpr2
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ $vgpr5 = V_ADD_U32_e32 1, $vgpr6, implicit $exec
+ S_ENDPGM 0
+...
+
+---
+name: iglp_input_coexec
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: name: iglp_input_coexec
+ ; CHECK: $vgpr2_vgpr3 = DS_READ_B64_TR_B16
+ ; CHECK: $vgpr4 = V_ADD_U32_e32 1,
+ bb.0:
+ liveins: $vgpr0, $vgpr4, $exec
+
+ IGLP_OPT 3
+ $vgpr2_vgpr3 = DS_READ_B64_TR_B16 $vgpr0, 64, 0, implicit $exec :: (load (s64), addrspace 3)
+ INLINEASM &"", attdialect, regdef:VReg_64_Align2, def $vgpr2_vgpr3, reguse tiedto:$0, $vgpr2_vgpr3(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 1, $vgpr4, implicit $exec
+ INLINEASM &"s_nop 0", sideeffect attdialect, regdef:VGPR_32, def $vgpr4, reguse tiedto:$0, $vgpr4(tied-def 3)
+ $vgpr4 = V_ADD_U32_e32 2, $vgpr4, implicit $exec
+ $vgpr5 = V_ADD_U32_e32 $vgpr2, $vgpr3, implicit $exec
+ S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/postra-inlineasm-def-latency.mir b/llvm/test/CodeGen/AMDGPU/postra-inlineasm-def-latency.mir
new file mode 100644
index 00000000000000..68c9ed918871f4
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/postra-inlineasm-def-latency.mir
@@ -0,0 +1,26 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 6
+# RUN: llc -mtriple=amdgpu9.42-amd-amdhsa -run-pass=postmisched,post-RA-hazard-rec -verify-misched %s -o - | FileCheck %s
+# RUN: llc -mtriple=amdgpu9.50-amd-amdhsa -run-pass=postmisched,post-RA-hazard-rec -verify-misched %s -o - | FileCheck %s
+
+# Keep an independent scalar move between an asm producer and consumer rather
+# than relying on the hazard recognizer to insert an S_NOP afterward.
+
+---
+name: asm_producer_latency
+tracksRegLiveness: true
+body: |
+ bb.0:
+ ; CHECK-LABEL: name: asm_producer_latency
+ ; CHECK: INLINEASM &"; def $0", attdialect, regdef:SGPR_192, def $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5
+ ; CHECK-NEXT: $sgpr10 = S_MOV_B32 killed $sgpr0
+ ; CHECK-NEXT: INLINEASM &"; def $0", attdialect, regdef:SGPR_192, def $sgpr4_sgpr5_sgpr6_sgpr7_sgpr8_sgpr9
+ ; CHECK-NEXT: $sgpr11 = S_MOV_B32 killed $sgpr1
+ ; CHECK-NEXT: INLINEASM &"; use $0", sideeffect attdialect, reguse:SGPR_128, killed $sgpr8_sgpr9_sgpr10_sgpr11
+ ; CHECK-NEXT: S_ENDPGM 0
+ INLINEASM &"; def $0", attdialect, regdef:SGPR_192, def $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5
+ INLINEASM &"; def $0", attdialect, regdef:SGPR_192, def $sgpr4_sgpr5_sgpr6_sgpr7_sgpr8_sgpr9
+ $sgpr10 = S_MOV_B32 $sgpr0
+ $sgpr11 = S_MOV_B32 $sgpr1
+ INLINEASM &"; use $0", sideeffect attdialect, reguse:SGPR_128, $sgpr8_sgpr9_sgpr10_sgpr11
+ S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll b/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll
index 53cbb4b26eed1a..4c454e636ac0da 100644
--- a/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll
+++ b/llvm/test/CodeGen/AMDGPU/preserve-wwm-copy-dst-reg.ll
@@ -152,42 +152,42 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
; GFX906-NEXT: s_mov_b64 exec, s[34:35]
; GFX906-NEXT: s_waitcnt vmcnt(0)
; GFX906-NEXT: v_readlane_b32 s11, v39, 12
+; GFX906-NEXT: v_readlane_b32 s12, v39, 13
+; GFX906-NEXT: v_readlane_b32 s13, v39, 14
+; GFX906-NEXT: v_readlane_b32 s14, v39, 15
+; GFX906-NEXT: v_readlane_b32 s15, v39, 16
+; GFX906-NEXT: v_readlane_b32 s16, v39, 17
+; GFX906-NEXT: v_readlane_b32 s17, v39, 18
+; GFX906-NEXT: v_readlane_b32 s18, v39, 19
+; GFX906-NEXT: v_readlane_b32 s19, v39, 20
+; GFX906-NEXT: v_readlane_b32 s20, v39, 21
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s11
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s12, v39, 13
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s12
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s13, v39, 14
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s13
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s14, v39, 15
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s14
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s15, v39, 16
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s15
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s16, v39, 17
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s16
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s17, v39, 18
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s17
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s18, v39, 19
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s18
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s19, v39, 20
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s19
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s20, v39, 21
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s20
; GFX906-NEXT: ;;#ASMEND
@@ -255,6 +255,15 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
; GFX906-NEXT: v_readlane_b32 s8, v39, 6
; GFX906-NEXT: v_readlane_b32 s10, v39, 4
; GFX906-NEXT: v_readlane_b32 s16, v39, 22
+; GFX906-NEXT: v_readlane_b32 s21, v39, 12
+; GFX906-NEXT: v_readlane_b32 s22, v39, 13
+; GFX906-NEXT: v_readlane_b32 s23, v39, 14
+; GFX906-NEXT: v_readlane_b32 s24, v39, 15
+; GFX906-NEXT: v_readlane_b32 s25, v39, 16
+; GFX906-NEXT: v_readlane_b32 s26, v39, 17
+; GFX906-NEXT: v_readlane_b32 s27, v39, 18
+; GFX906-NEXT: v_readlane_b32 s28, v39, 19
+; GFX906-NEXT: v_readlane_b32 s29, v39, 20
; GFX906-NEXT: v_readlane_b32 s5, v39, 11
; GFX906-NEXT: v_readlane_b32 s7, v39, 9
; GFX906-NEXT: v_readlane_b32 s9, v39, 7
@@ -265,39 +274,30 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
; GFX906-NEXT: v_readlane_b32 s15, v39, 0
; GFX906-NEXT: v_mov_b32_e32 v31, v40
; GFX906-NEXT: v_readlane_b32 s17, v39, 23
-; GFX906-NEXT: v_readlane_b32 s21, v39, 12
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s21
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s22, v39, 13
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s22
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s23, v39, 14
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s23
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s24, v39, 15
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s24
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s25, v39, 16
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s25
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s26, v39, 17
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s26
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s27, v39, 18
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s27
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s28, v39, 19
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s28
; GFX906-NEXT: ;;#ASMEND
-; GFX906-NEXT: v_readlane_b32 s29, v39, 20
; GFX906-NEXT: ;;#ASMSTART
; GFX906-NEXT: ; use s29
; GFX906-NEXT: ;;#ASMEND
@@ -532,42 +532,42 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
; GFX908-NEXT: s_mov_b64 exec, s[34:35]
; GFX908-NEXT: s_waitcnt vmcnt(0)
; GFX908-NEXT: v_readlane_b32 s11, v39, 12
+; GFX908-NEXT: v_readlane_b32 s12, v39, 13
+; GFX908-NEXT: v_readlane_b32 s13, v39, 14
+; GFX908-NEXT: v_readlane_b32 s14, v39, 15
+; GFX908-NEXT: v_readlane_b32 s15, v39, 16
+; GFX908-NEXT: v_readlane_b32 s16, v39, 17
+; GFX908-NEXT: v_readlane_b32 s17, v39, 18
+; GFX908-NEXT: v_readlane_b32 s18, v39, 19
+; GFX908-NEXT: v_readlane_b32 s19, v39, 20
+; GFX908-NEXT: v_readlane_b32 s20, v39, 21
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s11
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s12, v39, 13
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s12
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s13, v39, 14
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s13
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s14, v39, 15
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s14
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s15, v39, 16
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s15
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s16, v39, 17
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s16
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s17, v39, 18
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s17
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s18, v39, 19
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s18
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s19, v39, 20
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s19
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s20, v39, 21
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s20
; GFX908-NEXT: ;;#ASMEND
@@ -635,6 +635,15 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
; GFX908-NEXT: v_readlane_b32 s8, v39, 6
; GFX908-NEXT: v_readlane_b32 s10, v39, 4
; GFX908-NEXT: v_readlane_b32 s16, v39, 22
+; GFX908-NEXT: v_readlane_b32 s21, v39, 12
+; GFX908-NEXT: v_readlane_b32 s22, v39, 13
+; GFX908-NEXT: v_readlane_b32 s23, v39, 14
+; GFX908-NEXT: v_readlane_b32 s24, v39, 15
+; GFX908-NEXT: v_readlane_b32 s25, v39, 16
+; GFX908-NEXT: v_readlane_b32 s26, v39, 17
+; GFX908-NEXT: v_readlane_b32 s27, v39, 18
+; GFX908-NEXT: v_readlane_b32 s28, v39, 19
+; GFX908-NEXT: v_readlane_b32 s29, v39, 20
; GFX908-NEXT: v_readlane_b32 s5, v39, 11
; GFX908-NEXT: v_readlane_b32 s7, v39, 9
; GFX908-NEXT: v_readlane_b32 s9, v39, 7
@@ -645,39 +654,30 @@ define void @preserve_wwm_copy_dstreg(ptr %parg0, ptr %parg1, ptr %parg2) #0 {
; GFX908-NEXT: v_readlane_b32 s15, v39, 0
; GFX908-NEXT: v_mov_b32_e32 v31, v40
; GFX908-NEXT: v_readlane_b32 s17, v39, 23
-; GFX908-NEXT: v_readlane_b32 s21, v39, 12
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s21
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s22, v39, 13
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s22
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s23, v39, 14
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s23
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s24, v39, 15
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s24
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s25, v39, 16
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s25
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s26, v39, 17
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s26
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s27, v39, 18
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s27
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s28, v39, 19
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s28
; GFX908-NEXT: ;;#ASMEND
-; GFX908-NEXT: v_readlane_b32 s29, v39, 20
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ; use s29
; GFX908-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll b/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll
index 94cdcadaa68595..13a1881b8c5647 100644
--- a/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll
+++ b/llvm/test/CodeGen/AMDGPU/s_cmp_0.ll
@@ -533,8 +533,8 @@ define amdgpu_ps i32 @bcnt164(i64 inreg %val0) {
; CHECK-LABEL: bcnt164:
; CHECK: ; %bb.0:
; CHECK-NEXT: s_bcnt1_i32_b64 s0, s[0:1]
-; CHECK-NEXT: s_cmp_lg_u32 s0, 0
; CHECK-NEXT: s_mov_b32 s1, 0
+; CHECK-NEXT: s_cmp_lg_u32 s0, 0
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use s[0:1]
; CHECK-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll b/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll
index d06ea9b5b5f67d..ac3f4f8e71b2ef 100644
--- a/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll
+++ b/llvm/test/CodeGen/AMDGPU/scc-clobbered-sgpr-to-vmem-spill.ll
@@ -199,13 +199,13 @@ define amdgpu_kernel void @kernel0(ptr addrspace(1) %out, i32 %in) #1 {
; CHECK-NEXT: ; %bb.1: ; %ret
; CHECK-NEXT: s_endpgm
; CHECK-NEXT: .LBB0_2: ; %bb0
-; CHECK-NEXT: v_readlane_b32 s4, v22, 2
; CHECK-NEXT: s_mov_b64 s[16:17], s[0:1]
; CHECK-NEXT: v_readlane_b32 s0, v22, 0
+; CHECK-NEXT: v_readlane_b32 s4, v22, 2
+; CHECK-NEXT: v_readlane_b32 s1, v22, 1
; CHECK-NEXT: v_readlane_b32 s5, v22, 3
; CHECK-NEXT: v_readlane_b32 s6, v22, 4
; CHECK-NEXT: v_readlane_b32 s7, v22, 5
-; CHECK-NEXT: v_readlane_b32 s1, v22, 1
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use s[0:1]
; CHECK-NEXT: ;;#ASMEND
@@ -348,15 +348,6 @@ define amdgpu_kernel void @kernel0(ptr addrspace(1) %out, i32 %in) #1 {
; CHECK-NEXT: ; use s[0:15]
; CHECK-NEXT: ;;#ASMEND
; CHECK-NEXT: v_readlane_b32 s0, v23, 34
-; CHECK-NEXT: ;;#ASMSTART
-; CHECK-NEXT: ; use s[16:17]
-; CHECK-NEXT: ;;#ASMEND
-; CHECK-NEXT: ;;#ASMSTART
-; CHECK-NEXT: ; use s[20:23]
-; CHECK-NEXT: ;;#ASMEND
-; CHECK-NEXT: ;;#ASMSTART
-; CHECK-NEXT: ; use s[24:31]
-; CHECK-NEXT: ;;#ASMEND
; CHECK-NEXT: v_readlane_b32 s1, v23, 35
; CHECK-NEXT: v_readlane_b32 s2, v23, 36
; CHECK-NEXT: v_readlane_b32 s3, v23, 37
@@ -373,6 +364,15 @@ define amdgpu_kernel void @kernel0(ptr addrspace(1) %out, i32 %in) #1 {
; CHECK-NEXT: v_readlane_b32 s14, v23, 48
; CHECK-NEXT: v_readlane_b32 s15, v23, 49
; CHECK-NEXT: ;;#ASMSTART
+; CHECK-NEXT: ; use s[16:17]
+; CHECK-NEXT: ;;#ASMEND
+; CHECK-NEXT: ;;#ASMSTART
+; CHECK-NEXT: ; use s[20:23]
+; CHECK-NEXT: ;;#ASMEND
+; CHECK-NEXT: ;;#ASMSTART
+; CHECK-NEXT: ; use s[24:31]
+; CHECK-NEXT: ;;#ASMEND
+; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use s[0:15]
; CHECK-NEXT: ;;#ASMEND
; CHECK-NEXT: s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll b/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll
index 8963d38ef185c8..a859a5c5bcd694 100644
--- a/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/shufflevector.v2i64.v8i64.ll
@@ -14610,9 +14610,9 @@ define void @s_shuffle_v2i64_v8i64__15_2() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s30
; GFX900-NEXT: s_mov_b32 s9, s31
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -14639,9 +14639,9 @@ define void @s_shuffle_v2i64_v8i64__15_2() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s30
; GFX90A-NEXT: s_mov_b32 s9, s31
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -14750,9 +14750,9 @@ define void @s_shuffle_v2i64_v8i64__15_4() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s30
; GFX900-NEXT: s_mov_b32 s9, s31
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -14779,9 +14779,9 @@ define void @s_shuffle_v2i64_v8i64__15_4() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s30
; GFX90A-NEXT: s_mov_b32 s9, s31
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -14810,8 +14810,8 @@ define void @s_shuffle_v2i64_v8i64__15_4() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s30
; GFX942-NEXT: s_mov_b32 s9, s31
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -14845,9 +14845,9 @@ define void @s_shuffle_v2i64_v8i64__15_5() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s12, s30
; GFX900-NEXT: s_mov_b32 s13, s31
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -14874,9 +14874,9 @@ define void @s_shuffle_v2i64_v8i64__15_5() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s12, s30
; GFX90A-NEXT: s_mov_b32 s13, s31
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -15010,9 +15010,9 @@ define void @s_shuffle_v2i64_v8i64__15_6() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s30
; GFX942-NEXT: s_mov_b32 s9, s31
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s10, s12
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -15129,9 +15129,9 @@ define void @s_shuffle_v2i64_v8i64__15_7() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s12, s30
; GFX942-NEXT: s_mov_b32 s13, s31
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -15978,10 +15978,10 @@ define void @s_shuffle_v2i64_v8i64__10_0() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[4:19]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s10, s36
; GFX900-NEXT: s_mov_b32 s11, s37
@@ -16017,10 +16017,10 @@ define void @s_shuffle_v2i64_v8i64__10_0() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[4:19]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s10, s36
; GFX90A-NEXT: s_mov_b32 s11, s37
@@ -16179,8 +16179,8 @@ define void @s_shuffle_v2i64_v8i64__12_0() #0 {
; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s10, s16
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s11, s17
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -16708,10 +16708,10 @@ define void @s_shuffle_v2i64_v8i64__10_1() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[4:19]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s10, s38
; GFX900-NEXT: s_mov_b32 s11, s39
@@ -16747,10 +16747,10 @@ define void @s_shuffle_v2i64_v8i64__10_1() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[4:19]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s10, s38
; GFX90A-NEXT: s_mov_b32 s11, s39
@@ -16903,8 +16903,8 @@ define void @s_shuffle_v2i64_v8i64__12_1() #0 {
; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s10, s18
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s11, s19
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -17486,11 +17486,11 @@ define void @s_shuffle_v2i64_v8i64__9_2() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[16:31]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s18
; GFX900-NEXT: s_mov_b32 s9, s19
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -17515,11 +17515,11 @@ define void @s_shuffle_v2i64_v8i64__9_2() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[16:31]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s18
; GFX90A-NEXT: s_mov_b32 s9, s19
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -17570,9 +17570,9 @@ define void @s_shuffle_v2i64_v8i64__10_2() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s10, s20
; GFX900-NEXT: s_mov_b32 s11, s21
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -17597,9 +17597,9 @@ define void @s_shuffle_v2i64_v8i64__10_2() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s10, s20
; GFX90A-NEXT: s_mov_b32 s11, s21
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -17626,9 +17626,9 @@ define void @s_shuffle_v2i64_v8i64__10_2() #0 {
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; def s[4:19]
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s10, s20
; GFX942-NEXT: s_mov_b32 s11, s21
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -17660,11 +17660,11 @@ define void @s_shuffle_v2i64_v8i64__11_2() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[16:31]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s22
; GFX900-NEXT: s_mov_b32 s9, s23
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -17689,11 +17689,11 @@ define void @s_shuffle_v2i64_v8i64__11_2() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[16:31]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s22
; GFX90A-NEXT: s_mov_b32 s9, s23
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -17804,11 +17804,11 @@ define void @s_shuffle_v2i64_v8i64__13_2() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[16:31]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s26
; GFX900-NEXT: s_mov_b32 s9, s27
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -17833,11 +17833,11 @@ define void @s_shuffle_v2i64_v8i64__13_2() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[16:31]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s26
; GFX90A-NEXT: s_mov_b32 s9, s27
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -18321,9 +18321,9 @@ define void @s_shuffle_v2i64_v8i64__10_3() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s10, s22
; GFX900-NEXT: s_mov_b32 s11, s23
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -18348,9 +18348,9 @@ define void @s_shuffle_v2i64_v8i64__10_3() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s10, s22
; GFX90A-NEXT: s_mov_b32 s11, s23
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -18377,9 +18377,9 @@ define void @s_shuffle_v2i64_v8i64__10_3() #0 {
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; def s[4:19]
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s10, s22
; GFX942-NEXT: s_mov_b32 s11, s23
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -18957,11 +18957,11 @@ define void @s_shuffle_v2i64_v8i64__9_4() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[16:31]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s18
; GFX900-NEXT: s_mov_b32 s9, s19
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -18986,11 +18986,11 @@ define void @s_shuffle_v2i64_v8i64__9_4() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[16:31]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s18
; GFX90A-NEXT: s_mov_b32 s9, s19
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -19018,9 +19018,9 @@ define void @s_shuffle_v2i64_v8i64__9_4() #0 {
; GFX942-NEXT: ; def s[16:31]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s18
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s19
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -19108,11 +19108,11 @@ define void @s_shuffle_v2i64_v8i64__11_4() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[16:31]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s22
; GFX900-NEXT: s_mov_b32 s9, s23
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -19137,11 +19137,11 @@ define void @s_shuffle_v2i64_v8i64__11_4() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[16:31]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s22
; GFX90A-NEXT: s_mov_b32 s9, s23
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -19169,9 +19169,9 @@ define void @s_shuffle_v2i64_v8i64__11_4() #0 {
; GFX942-NEXT: ; def s[16:31]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s22
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s23
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -19205,9 +19205,9 @@ define void @s_shuffle_v2i64_v8i64__12_4() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s26, s12
; GFX900-NEXT: s_mov_b32 s27, s13
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -19234,9 +19234,9 @@ define void @s_shuffle_v2i64_v8i64__12_4() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s26, s12
; GFX90A-NEXT: s_mov_b32 s27, s13
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -19285,11 +19285,11 @@ define void @s_shuffle_v2i64_v8i64__13_4() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[16:31]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s26
; GFX900-NEXT: s_mov_b32 s9, s27
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -19314,11 +19314,11 @@ define void @s_shuffle_v2i64_v8i64__13_4() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[16:31]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s26
; GFX90A-NEXT: s_mov_b32 s9, s27
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -19346,9 +19346,9 @@ define void @s_shuffle_v2i64_v8i64__13_4() #0 {
; GFX942-NEXT: ; def s[16:31]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s26
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s27
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -19883,9 +19883,9 @@ define void @s_shuffle_v2i64_v8i64__9_5() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s12, s18
; GFX900-NEXT: s_mov_b32 s13, s19
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -19912,9 +19912,9 @@ define void @s_shuffle_v2i64_v8i64__9_5() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s12, s18
; GFX90A-NEXT: s_mov_b32 s13, s19
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -20021,9 +20021,9 @@ define void @s_shuffle_v2i64_v8i64__11_5() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s12, s22
; GFX900-NEXT: s_mov_b32 s13, s23
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -20050,9 +20050,9 @@ define void @s_shuffle_v2i64_v8i64__11_5() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s12, s22
; GFX90A-NEXT: s_mov_b32 s13, s23
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -20103,9 +20103,9 @@ define void @s_shuffle_v2i64_v8i64__12_5() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s26, s14
; GFX900-NEXT: s_mov_b32 s27, s15
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -20132,9 +20132,9 @@ define void @s_shuffle_v2i64_v8i64__12_5() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s26, s14
; GFX90A-NEXT: s_mov_b32 s27, s15
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -20185,9 +20185,9 @@ define void @s_shuffle_v2i64_v8i64__13_5() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s12, s26
; GFX900-NEXT: s_mov_b32 s13, s27
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -20214,9 +20214,9 @@ define void @s_shuffle_v2i64_v8i64__13_5() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s12, s26
; GFX90A-NEXT: s_mov_b32 s13, s27
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -20859,16 +20859,16 @@ define void @s_shuffle_v2i64_v8i64__9_6() #0 {
; GFX942-NEXT: s_nop 1
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
; GFX942-NEXT: ;;#ASMSTART
-; GFX942-NEXT: ; def s[16:31]
+; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: ;;#ASMSTART
-; GFX942-NEXT: ; def s[0:15]
+; GFX942-NEXT: ; def s[16:31]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s18
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s19
; GFX942-NEXT: s_mov_b32 s10, s12
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -21034,16 +21034,16 @@ define void @s_shuffle_v2i64_v8i64__11_6() #0 {
; GFX942-NEXT: s_nop 1
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
; GFX942-NEXT: ;;#ASMSTART
-; GFX942-NEXT: ; def s[16:31]
+; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: ;;#ASMSTART
-; GFX942-NEXT: ; def s[0:15]
+; GFX942-NEXT: ; def s[16:31]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s22
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s23
; GFX942-NEXT: s_mov_b32 s10, s12
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -21259,16 +21259,16 @@ define void @s_shuffle_v2i64_v8i64__13_6() #0 {
; GFX942-NEXT: s_nop 1
; GFX942-NEXT: v_writelane_b32 v0, s31, 1
; GFX942-NEXT: ;;#ASMSTART
-; GFX942-NEXT: ; def s[16:31]
+; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: ;;#ASMSTART
-; GFX942-NEXT: ; def s[0:15]
+; GFX942-NEXT: ; def s[16:31]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s26
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s27
; GFX942-NEXT: s_mov_b32 s10, s12
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -21930,9 +21930,9 @@ define void @s_shuffle_v2i64_v8i64__9_7() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s12, s18
; GFX942-NEXT: s_mov_b32 s13, s19
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -22105,9 +22105,9 @@ define void @s_shuffle_v2i64_v8i64__11_7() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s12, s22
; GFX942-NEXT: s_mov_b32 s13, s23
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -22330,9 +22330,9 @@ define void @s_shuffle_v2i64_v8i64__13_7() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s12, s26
; GFX942-NEXT: s_mov_b32 s13, s27
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -23348,10 +23348,10 @@ define void @s_shuffle_v2i64_v8i64__3_9() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[8:23]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[8:23]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s42
; GFX900-NEXT: s_mov_b32 s9, s43
@@ -23387,10 +23387,10 @@ define void @s_shuffle_v2i64_v8i64__3_9() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[8:23]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[8:23]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s42
; GFX90A-NEXT: s_mov_b32 s9, s43
@@ -23451,9 +23451,9 @@ define void @s_shuffle_v2i64_v8i64__4_9() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s18
; GFX900-NEXT: s_mov_b32 s15, s19
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -23480,9 +23480,9 @@ define void @s_shuffle_v2i64_v8i64__4_9() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s18
; GFX90A-NEXT: s_mov_b32 s15, s19
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -23531,9 +23531,9 @@ define void @s_shuffle_v2i64_v8i64__5_9() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[8:23]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s26
; GFX900-NEXT: s_mov_b32 s9, s27
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -23558,9 +23558,9 @@ define void @s_shuffle_v2i64_v8i64__5_9() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[8:23]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s26
; GFX90A-NEXT: s_mov_b32 s9, s27
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -23587,9 +23587,9 @@ define void @s_shuffle_v2i64_v8i64__5_9() #0 {
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; def s[8:23]
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s8, s26
; GFX942-NEXT: s_mov_b32 s9, s27
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -23706,9 +23706,9 @@ define void @s_shuffle_v2i64_v8i64__6_9() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s18
; GFX942-NEXT: s_mov_b32 s15, s19
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -24159,10 +24159,10 @@ define void @s_shuffle_v2i64_v8i64__1_10() #0 {
; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s18
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s19
; GFX942-NEXT: s_mov_b32 s10, s4
; GFX942-NEXT: s_mov_b32 s11, s5
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -24251,10 +24251,10 @@ define void @s_shuffle_v2i64_v8i64__3_10() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[8:23]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[8:23]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s42
; GFX900-NEXT: s_mov_b32 s9, s43
@@ -24292,10 +24292,10 @@ define void @s_shuffle_v2i64_v8i64__3_10() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[8:23]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[8:23]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s42
; GFX90A-NEXT: s_mov_b32 s9, s43
@@ -24360,9 +24360,9 @@ define void @s_shuffle_v2i64_v8i64__4_10() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s20
; GFX900-NEXT: s_mov_b32 s15, s21
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -24389,9 +24389,9 @@ define void @s_shuffle_v2i64_v8i64__4_10() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s20
; GFX90A-NEXT: s_mov_b32 s15, s21
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -24440,11 +24440,11 @@ define void @s_shuffle_v2i64_v8i64__5_10() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[8:23]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s26
; GFX900-NEXT: s_mov_b32 s9, s27
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -24469,11 +24469,11 @@ define void @s_shuffle_v2i64_v8i64__5_10() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[8:23]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s26
; GFX90A-NEXT: s_mov_b32 s9, s27
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -24609,9 +24609,9 @@ define void @s_shuffle_v2i64_v8i64__6_10() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s20
; GFX942-NEXT: s_mov_b32 s15, s21
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -25101,10 +25101,10 @@ define void @s_shuffle_v2i64_v8i64__1_11() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[4:19]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s38
; GFX900-NEXT: s_mov_b32 s9, s39
@@ -25140,10 +25140,10 @@ define void @s_shuffle_v2i64_v8i64__1_11() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[4:19]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s38
; GFX90A-NEXT: s_mov_b32 s9, s39
@@ -25258,9 +25258,9 @@ define void @s_shuffle_v2i64_v8i64__3_11() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s22
; GFX900-NEXT: s_mov_b32 s9, s23
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -25285,9 +25285,9 @@ define void @s_shuffle_v2i64_v8i64__3_11() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s22
; GFX90A-NEXT: s_mov_b32 s9, s23
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -25314,9 +25314,9 @@ define void @s_shuffle_v2i64_v8i64__3_11() #0 {
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; def s[4:19]
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s8, s22
; GFX942-NEXT: s_mov_b32 s9, s23
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -25350,9 +25350,9 @@ define void @s_shuffle_v2i64_v8i64__4_11() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s22
; GFX900-NEXT: s_mov_b32 s15, s23
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -25379,9 +25379,9 @@ define void @s_shuffle_v2i64_v8i64__4_11() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s22
; GFX90A-NEXT: s_mov_b32 s15, s23
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -25571,9 +25571,9 @@ define void @s_shuffle_v2i64_v8i64__6_11() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s22
; GFX942-NEXT: s_mov_b32 s15, s23
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -25987,10 +25987,10 @@ define void @s_shuffle_v2i64_v8i64__1_12() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[4:19]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s38
; GFX900-NEXT: s_mov_b32 s9, s39
@@ -26028,10 +26028,10 @@ define void @s_shuffle_v2i64_v8i64__1_12() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[4:19]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s38
; GFX90A-NEXT: s_mov_b32 s9, s39
@@ -26150,11 +26150,11 @@ define void @s_shuffle_v2i64_v8i64__3_12() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s22
; GFX900-NEXT: s_mov_b32 s9, s23
; GFX900-NEXT: s_mov_b32 s10, s12
; GFX900-NEXT: s_mov_b32 s11, s13
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -26179,11 +26179,11 @@ define void @s_shuffle_v2i64_v8i64__3_12() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s22
; GFX90A-NEXT: s_mov_b32 s9, s23
; GFX90A-NEXT: s_mov_b32 s10, s12
; GFX90A-NEXT: s_mov_b32 s11, s13
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -26210,11 +26210,11 @@ define void @s_shuffle_v2i64_v8i64__3_12() #0 {
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; def s[4:19]
; GFX942-NEXT: ;;#ASMEND
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s8, s22
; GFX942-NEXT: s_mov_b32 s9, s23
; GFX942-NEXT: s_mov_b32 s10, s12
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -26248,9 +26248,9 @@ define void @s_shuffle_v2i64_v8i64__4_12() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s24
; GFX900-NEXT: s_mov_b32 s15, s25
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -26277,9 +26277,9 @@ define void @s_shuffle_v2i64_v8i64__4_12() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s24
; GFX90A-NEXT: s_mov_b32 s15, s25
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -26475,9 +26475,9 @@ define void @s_shuffle_v2i64_v8i64__6_12() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s24
; GFX942-NEXT: s_mov_b32 s15, s25
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -26918,8 +26918,8 @@ define void @s_shuffle_v2i64_v8i64__1_13() #0 {
; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s18
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s19
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -27069,9 +27069,9 @@ define void @s_shuffle_v2i64_v8i64__4_13() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s26
; GFX900-NEXT: s_mov_b32 s15, s27
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -27098,9 +27098,9 @@ define void @s_shuffle_v2i64_v8i64__4_13() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s26
; GFX90A-NEXT: s_mov_b32 s15, s27
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -27151,9 +27151,9 @@ define void @s_shuffle_v2i64_v8i64__5_13() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s24, s14
; GFX900-NEXT: s_mov_b32 s25, s15
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[24:25]
; GFX900-NEXT: s_mov_b64 s[10:11], s[26:27]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -27180,9 +27180,9 @@ define void @s_shuffle_v2i64_v8i64__5_13() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s24, s14
; GFX90A-NEXT: s_mov_b32 s25, s15
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[24:25]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[26:27]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -27316,9 +27316,9 @@ define void @s_shuffle_v2i64_v8i64__6_13() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s26
; GFX942-NEXT: s_mov_b32 s15, s27
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -27846,10 +27846,10 @@ define void @s_shuffle_v2i64_v8i64__1_14() #0 {
; GFX900-NEXT: v_writelane_b32 v0, s50, 6
; GFX900-NEXT: v_writelane_b32 v0, s51, 7
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[36:51]
+; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: ;;#ASMSTART
-; GFX900-NEXT: ; def s[4:19]
+; GFX900-NEXT: ; def s[36:51]
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s8, s38
; GFX900-NEXT: s_mov_b32 s9, s39
@@ -27887,10 +27887,10 @@ define void @s_shuffle_v2i64_v8i64__1_14() #0 {
; GFX90A-NEXT: v_writelane_b32 v0, s50, 6
; GFX90A-NEXT: v_writelane_b32 v0, s51, 7
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[36:51]
+; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: ;;#ASMSTART
-; GFX90A-NEXT: ; def s[4:19]
+; GFX90A-NEXT: ; def s[36:51]
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s8, s38
; GFX90A-NEXT: s_mov_b32 s9, s39
@@ -27929,10 +27929,10 @@ define void @s_shuffle_v2i64_v8i64__1_14() #0 {
; GFX942-NEXT: ; def s[0:15]
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s8, s18
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b32 s9, s19
; GFX942-NEXT: s_mov_b32 s10, s12
; GFX942-NEXT: s_mov_b32 s11, s13
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -28020,11 +28020,11 @@ define void @s_shuffle_v2i64_v8i64__3_14() #0 {
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; def s[4:19]
; GFX900-NEXT: ;;#ASMEND
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b32 s8, s22
; GFX900-NEXT: s_mov_b32 s9, s23
; GFX900-NEXT: s_mov_b32 s10, s16
; GFX900-NEXT: s_mov_b32 s11, s17
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -28049,11 +28049,11 @@ define void @s_shuffle_v2i64_v8i64__3_14() #0 {
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; def s[4:19]
; GFX90A-NEXT: ;;#ASMEND
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b32 s8, s22
; GFX90A-NEXT: s_mov_b32 s9, s23
; GFX90A-NEXT: s_mov_b32 s10, s16
; GFX90A-NEXT: s_mov_b32 s11, s17
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -28106,9 +28106,9 @@ define void @s_shuffle_v2i64_v8i64__4_14() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s28
; GFX900-NEXT: s_mov_b32 s15, s29
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -28135,9 +28135,9 @@ define void @s_shuffle_v2i64_v8i64__4_14() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s28
; GFX90A-NEXT: s_mov_b32 s15, s29
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -28333,9 +28333,9 @@ define void @s_shuffle_v2i64_v8i64__6_14() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s28
; GFX942-NEXT: s_mov_b32 s15, s29
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
@@ -29011,9 +29011,9 @@ define void @s_shuffle_v2i64_v8i64__4_15() #0 {
; GFX900-NEXT: ;;#ASMEND
; GFX900-NEXT: s_mov_b32 s14, s30
; GFX900-NEXT: s_mov_b32 s15, s31
-; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX900-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX900-NEXT: v_readlane_b32 s30, v0, 0
; GFX900-NEXT: ;;#ASMSTART
; GFX900-NEXT: ; use s[8:11]
; GFX900-NEXT: ;;#ASMEND
@@ -29040,9 +29040,9 @@ define void @s_shuffle_v2i64_v8i64__4_15() #0 {
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_mov_b32 s14, s30
; GFX90A-NEXT: s_mov_b32 s15, s31
-; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX90A-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX90A-NEXT: v_readlane_b32 s30, v0, 0
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ; use s[8:11]
; GFX90A-NEXT: ;;#ASMEND
@@ -29260,9 +29260,9 @@ define void @s_shuffle_v2i64_v8i64__6_15() #0 {
; GFX942-NEXT: ;;#ASMEND
; GFX942-NEXT: s_mov_b32 s14, s30
; GFX942-NEXT: s_mov_b32 s15, s31
-; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: s_mov_b64 s[8:9], s[12:13]
; GFX942-NEXT: s_mov_b64 s[10:11], s[14:15]
+; GFX942-NEXT: v_readlane_b32 s30, v0, 0
; GFX942-NEXT: ;;#ASMSTART
; GFX942-NEXT: ; use s[8:11]
; GFX942-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/spill-agpr.ll b/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
index 1ca643d90aac02..364ccdde2002ea 100644
--- a/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
+++ b/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
@@ -34,11 +34,11 @@ define amdgpu_kernel void @max_12regs_13a_used(i32 %cond, ptr addrspace(1) %arg,
; GFX908-NEXT: v_accvgpr_read_b32 v2, a5
; GFX908-NEXT: v_accvgpr_read_b32 v3, a6
; GFX908-NEXT: v_accvgpr_read_b32 v4, a7
-; GFX908-NEXT: v_accvgpr_write_b32 a4, 4
; GFX908-NEXT: v_accvgpr_write_b32 a8, 5
; GFX908-NEXT: v_accvgpr_write_b32 a9, 1
; GFX908-NEXT: v_accvgpr_write_b32 a10, 2
; GFX908-NEXT: v_accvgpr_write_b32 a11, 3
+; GFX908-NEXT: v_accvgpr_write_b32 a4, 4
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ;;#ASMEND
; GFX908-NEXT: v_accvgpr_write_b32 a7, v4
@@ -77,17 +77,16 @@ define amdgpu_kernel void @max_12regs_13a_used(i32 %cond, ptr addrspace(1) %arg,
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: s_endpgm
; GFX90A-NEXT: .LBB0_2: ; %use
-; GFX90A-NEXT: s_nop 3
+; GFX90A-NEXT: v_accvgpr_write_b32 a8, 5
+; GFX90A-NEXT: v_accvgpr_write_b32 a9, 1
+; GFX90A-NEXT: v_accvgpr_write_b32 a10, 2
+; GFX90A-NEXT: v_accvgpr_write_b32 a11, 3
; GFX90A-NEXT: v_accvgpr_read_b32 v9, a7
; GFX90A-NEXT: v_accvgpr_read_b32 v8, a6
; GFX90A-NEXT: v_accvgpr_read_b32 v7, a5
; GFX90A-NEXT: v_accvgpr_read_b32 v6, a4
; GFX90A-NEXT: v_accvgpr_write_b32 a4, 4
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
-; GFX90A-NEXT: v_accvgpr_write_b32 a8, 5
-; GFX90A-NEXT: v_accvgpr_write_b32 a9, 1
-; GFX90A-NEXT: v_accvgpr_write_b32 a10, 2
-; GFX90A-NEXT: v_accvgpr_write_b32 a11, 3
; GFX90A-NEXT: ;;#ASMSTART
; GFX90A-NEXT: ;;#ASMEND
; GFX90A-NEXT: v_accvgpr_write_b32 a4, v6
@@ -139,13 +138,14 @@ define amdgpu_kernel void @max_10_vgprs_used_9a() #1 {
; GFX908-NEXT: ;;#ASMEND
; GFX908-NEXT: v_accvgpr_read_b32 v5, a2
; GFX908-NEXT: v_accvgpr_read_b32 v4, a1
-; GFX908-NEXT: v_accvgpr_write_b32 a1, v0
; GFX908-NEXT: v_accvgpr_write_b32 a4, v3
; GFX908-NEXT: v_accvgpr_write_b32 a3, v2
; GFX908-NEXT: v_accvgpr_write_b32 a2, v1
+; GFX908-NEXT: v_accvgpr_write_b32 a1, v0
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ;;#ASMEND
; GFX908-NEXT: v_accvgpr_write_b32 a0, v4
+; GFX908-NEXT: s_nop 1
; GFX908-NEXT: v_accvgpr_write_b32 a1, v5
; GFX908-NEXT: ;;#ASMSTART
; GFX908-NEXT: ;;#ASMEND
diff --git a/llvm/test/CodeGen/AMDGPU/sub.ll b/llvm/test/CodeGen/AMDGPU/sub.ll
index 8a5f979568b0f6..851cb93cb68328 100644
--- a/llvm/test/CodeGen/AMDGPU/sub.ll
+++ b/llvm/test/CodeGen/AMDGPU/sub.ll
@@ -911,11 +911,11 @@ define amdgpu_kernel void @v_test_sub_v4i64(ptr addrspace(1) %out, ptr addrspace
define amdgpu_ps void @sub_select_vop3(i32 inreg %s, i32 %v) {
; GFX6-LABEL: sub_select_vop3:
; GFX6: ; %bb.0:
-; GFX6-NEXT: v_subrev_i32_e64 v0, s[0:1], s0, v0
-; GFX6-NEXT: s_mov_b32 m0, -1
; GFX6-NEXT: ;;#ASMSTART
; GFX6-NEXT: ; def vcc
; GFX6-NEXT: ;;#ASMEND
+; GFX6-NEXT: v_subrev_i32_e64 v0, s[0:1], s0, v0
+; GFX6-NEXT: s_mov_b32 m0, -1
; GFX6-NEXT: ds_write_b32 v0, v0
; GFX6-NEXT: ;;#ASMSTART
; GFX6-NEXT: ; use vcc
@@ -924,11 +924,11 @@ define amdgpu_ps void @sub_select_vop3(i32 inreg %s, i32 %v) {
;
; GFX8-LABEL: sub_select_vop3:
; GFX8: ; %bb.0:
-; GFX8-NEXT: v_subrev_u32_e64 v0, s[0:1], s0, v0
-; GFX8-NEXT: s_mov_b32 m0, -1
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; def vcc
; GFX8-NEXT: ;;#ASMEND
+; GFX8-NEXT: v_subrev_u32_e64 v0, s[0:1], s0, v0
+; GFX8-NEXT: s_mov_b32 m0, -1
; GFX8-NEXT: ds_write_b32 v0, v0
; GFX8-NEXT: ;;#ASMSTART
; GFX8-NEXT: ; use vcc
@@ -937,10 +937,10 @@ define amdgpu_ps void @sub_select_vop3(i32 inreg %s, i32 %v) {
;
; GFX9-LABEL: sub_select_vop3:
; GFX9: ; %bb.0:
-; GFX9-NEXT: v_subrev_u32_e32 v0, s0, v0
; GFX9-NEXT: ;;#ASMSTART
; GFX9-NEXT: ; def vcc
; GFX9-NEXT: ;;#ASMEND
+; GFX9-NEXT: v_subrev_u32_e32 v0, s0, v0
; GFX9-NEXT: ds_write_b32 v0, v0
; GFX9-NEXT: ;;#ASMSTART
; GFX9-NEXT: ; use vcc
diff --git a/llvm/test/CodeGen/AMDGPU/undefined-subreg-liverange.ll b/llvm/test/CodeGen/AMDGPU/undefined-subreg-liverange.ll
index 64bcad7003bd5d..e0bffeca609b8b 100644
--- a/llvm/test/CodeGen/AMDGPU/undefined-subreg-liverange.ll
+++ b/llvm/test/CodeGen/AMDGPU/undefined-subreg-liverange.ll
@@ -92,14 +92,14 @@ bb11: ; preds = %bb9
define amdgpu_kernel void @partially_undef_copy() #0 {
; CHECK-LABEL: partially_undef_copy:
; CHECK: ; %bb.0:
-; CHECK-NEXT: s_mov_b32 s3, 0xf000
-; CHECK-NEXT: s_mov_b32 s2, -1
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: v_mov_b32_e32 v5, 5
; CHECK-NEXT: ;;#ASMEND
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: v_mov_b32_e32 v6, 6
; CHECK-NEXT: ;;#ASMEND
+; CHECK-NEXT: s_mov_b32 s3, 0xf000
+; CHECK-NEXT: s_mov_b32 s2, -1
; CHECK-NEXT: buffer_store_dwordx4 v[6:9], off, s[0:3], 0
; CHECK-NEXT: s_waitcnt vmcnt(0)
; CHECK-NEXT: ;;#ASMSTART
diff --git a/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll b/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll
index 2550f4d1fbc85d..4428cd60c9bd02 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma-scale-src-perm.ll
@@ -65,17 +65,17 @@ define amdgpu_ps <8 x float> @wmma_scale_src_extra_use(<16 x i32> %a, <16 x i32>
; CHECK-NEXT: v_nop
; CHECK-NEXT: global_prefetch_b8 v0, s[64:65] scope:SCOPE_SE
; CHECK-NEXT: v_perm_b32 v40, s0, s1, 4
+; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1)
+; CHECK-NEXT: v_wmma_scale_f32_16x16x128_f8f6f4 v[32:39], v[0:15], v[16:31], v[32:39], v40, v40 matrix_a_fmt:MATRIX_FMT_BF8 matrix_b_fmt:MATRIX_FMT_BF8
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use v40
; CHECK-NEXT: ;;#ASMEND
-; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(TRANS32_DEP_1)
-; CHECK-NEXT: v_wmma_scale_f32_16x16x128_f8f6f4 v[32:39], v[0:15], v[16:31], v[32:39], v40, v40 matrix_a_fmt:MATRIX_FMT_BF8 matrix_b_fmt:MATRIX_FMT_BF8
; CHECK-NEXT: v_nop
; CHECK-NEXT: v_nop
; CHECK-NEXT: v_nop
; CHECK-NEXT: v_nop
-; CHECK-NEXT: v_dual_mov_b32 v0, v32 :: v_dual_mov_b32 v1, v33
; CHECK-NEXT: s_delay_alu instid0(TRANS32_DEP_1) | instskip(NEXT) | instid1(TRANS32_DEP_1)
+; CHECK-NEXT: v_dual_mov_b32 v0, v32 :: v_dual_mov_b32 v1, v33
; CHECK-NEXT: v_dual_mov_b32 v2, v34 :: v_dual_mov_b32 v3, v35
; CHECK-NEXT: v_dual_mov_b32 v4, v36 :: v_dual_mov_b32 v5, v37
; CHECK-NEXT: v_dual_mov_b32 v6, v38 :: v_dual_mov_b32 v7, v39
diff --git a/llvm/test/CodeGen/AMDGPU/write_register.ll b/llvm/test/CodeGen/AMDGPU/write_register.ll
index 6812d557b6769d..a5e56cedda8a69 100644
--- a/llvm/test/CodeGen/AMDGPU/write_register.ll
+++ b/llvm/test/CodeGen/AMDGPU/write_register.ll
@@ -222,13 +222,13 @@ define void @write_imm_into_src_flat_scratch_base() {
; CHECK-NEXT: s_wait_loadcnt_dscnt 0x0
; CHECK-NEXT: s_wait_kmcnt 0x0
; CHECK-NEXT: s_mov_b64 s[0:1], src_flat_scratch_base_lo
+; CHECK-NEXT: s_mov_b64 src_flat_scratch_base_lo, 0
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use s[0:1]
; CHECK-NEXT: ;;#ASMEND
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use s[0:1]
; CHECK-NEXT: ;;#ASMEND
-; CHECK-NEXT: s_mov_b64 src_flat_scratch_base_lo, 0
; CHECK-NEXT: s_mov_b64 src_flat_scratch_base_lo, -1
; CHECK-NEXT: s_set_pc_i64 s[30:31]
call void @llvm.write_register.i64(metadata !7, i64 0)
@@ -858,10 +858,10 @@ define void @write_sgpr_into_src_flat_scratch_base(i64 inreg %x) {
; CHECK-NEXT: s_wait_loadcnt_dscnt 0x0
; CHECK-NEXT: s_wait_kmcnt 0x0
; CHECK-NEXT: s_mov_b64 s[2:3], src_flat_scratch_base_lo
+; CHECK-NEXT: s_mov_b64 src_flat_scratch_base_lo, s[0:1]
; CHECK-NEXT: ;;#ASMSTART
; CHECK-NEXT: ; use s[2:3]
; CHECK-NEXT: ;;#ASMEND
-; CHECK-NEXT: s_mov_b64 src_flat_scratch_base_lo, s[0:1]
; CHECK-NEXT: s_set_pc_i64 s[30:31]
call void @llvm.write_register.i64(metadata !7, i64 %x)
%read = call i64 @llvm.read_register.i64(metadata !7)
diff --git a/llvm/test/CodeGen/PowerPC/inline-asm-physical-fpr-spe.ll b/llvm/test/CodeGen/PowerPC/inline-asm-physical-fpr-spe.ll
index 5c3ba3c2a6eac2..385df3d0a478d5 100644
--- a/llvm/test/CodeGen/PowerPC/inline-asm-physical-fpr-spe.ll
+++ b/llvm/test/CodeGen/PowerPC/inline-asm-physical-fpr-spe.ll
@@ -9,8 +9,8 @@ define i32 @test_f32(float %x) {
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset r31, -4
; CHECK-NEXT: stw 3, 8(1)
-; CHECK-NEXT: stw 31, 12(1) # 4-byte Folded Spill
; CHECK-NEXT: lwz 3, 8(1)
+; CHECK-NEXT: stw 31, 12(1) # 4-byte Folded Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: efsctsi 31, 3
; CHECK-NEXT: #NO_APP
diff --git a/llvm/test/CodeGen/PowerPC/not-fixed-frame-object.ll b/llvm/test/CodeGen/PowerPC/not-fixed-frame-object.ll
index e7778e7a4aa5da..175defbfc0befe 100644
--- a/llvm/test/CodeGen/PowerPC/not-fixed-frame-object.ll
+++ b/llvm/test/CodeGen/PowerPC/not-fixed-frame-object.ll
@@ -46,10 +46,10 @@ define dso_local signext i32 @caller(i32 signext %a, i32 signext %b, i32 signext
; CHECK-NEXT: std r31, -8(r1) # 8-byte Folded Spill
; CHECK-NEXT: stdu r1, -192(r1)
; CHECK-NEXT: std r0, 208(r1)
-; CHECK-NEXT: std r5, 32(r1) # 8-byte Folded Spill
-; CHECK-NEXT: mr r0, r4
; CHECK-NEXT: std r3, 40(r1) # 8-byte Folded Spill
+; CHECK-NEXT: mr r0, r4
; CHECK-NEXT: ld r3, 40(r1) # 8-byte Folded Reload
+; CHECK-NEXT: std r5, 32(r1) # 8-byte Folded Spill
; CHECK-NEXT: #APP
; CHECK-NEXT: add r3, r3, r0
; CHECK-NEXT: #NO_APP
diff --git a/llvm/test/CodeGen/RISCV/load-store-pair.ll b/llvm/test/CodeGen/RISCV/load-store-pair.ll
index 0413e9291dc889..42c91fbea6e468 100644
--- a/llvm/test/CodeGen/RISCV/load-store-pair.ll
+++ b/llvm/test/CodeGen/RISCV/load-store-pair.ll
@@ -33,10 +33,10 @@ define void @testi(ptr %a) {
; RV32I-NEXT: .cfi_offset s3, -8
; RV32I-NEXT: .cfi_offset s4, -12
; RV32I-NEXT: .cfi_offset s5, -16
-; RV32I-NEXT: lw s2, 4(a0)
; RV32I-NEXT: lw s3, 0(a0)
-; RV32I-NEXT: lw s4, 12(a0)
+; RV32I-NEXT: lw s2, 4(a0)
; RV32I-NEXT: lw s5, 8(a0)
+; RV32I-NEXT: lw s4, 12(a0)
; RV32I-NEXT: #APP
; RV32I-NEXT: #NO_APP
; RV32I-NEXT: lw s2, 12(sp) # 4-byte Folded Reload
@@ -63,10 +63,10 @@ define void @testi(ptr %a) {
; RV32D-NEXT: .cfi_offset s3, -8
; RV32D-NEXT: .cfi_offset s4, -12
; RV32D-NEXT: .cfi_offset s5, -16
-; RV32D-NEXT: lw s2, 4(a0)
; RV32D-NEXT: lw s3, 0(a0)
-; RV32D-NEXT: lw s4, 12(a0)
+; RV32D-NEXT: lw s2, 4(a0)
; RV32D-NEXT: lw s5, 8(a0)
+; RV32D-NEXT: lw s4, 12(a0)
; RV32D-NEXT: #APP
; RV32D-NEXT: #NO_APP
; RV32D-NEXT: lw s2, 12(sp) # 4-byte Folded Reload
@@ -93,10 +93,10 @@ define void @testi(ptr %a) {
; RV64I-NEXT: .cfi_offset s3, -16
; RV64I-NEXT: .cfi_offset s4, -24
; RV64I-NEXT: .cfi_offset s5, -32
-; RV64I-NEXT: ld s2, 8(a0)
; RV64I-NEXT: ld s3, 0(a0)
-; RV64I-NEXT: ld s4, 24(a0)
+; RV64I-NEXT: ld s2, 8(a0)
; RV64I-NEXT: ld s5, 16(a0)
+; RV64I-NEXT: ld s4, 24(a0)
; RV64I-NEXT: #APP
; RV64I-NEXT: #NO_APP
; RV64I-NEXT: ld s2, 24(sp) # 8-byte Folded Reload
@@ -123,10 +123,10 @@ define void @testi(ptr %a) {
; RV64D-NEXT: .cfi_offset s3, -16
; RV64D-NEXT: .cfi_offset s4, -24
; RV64D-NEXT: .cfi_offset s5, -32
-; RV64D-NEXT: ld s2, 8(a0)
; RV64D-NEXT: ld s3, 0(a0)
-; RV64D-NEXT: ld s4, 24(a0)
+; RV64D-NEXT: ld s2, 8(a0)
; RV64D-NEXT: ld s5, 16(a0)
+; RV64D-NEXT: ld s4, 24(a0)
; RV64D-NEXT: #APP
; RV64D-NEXT: #NO_APP
; RV64D-NEXT: ld s2, 24(sp) # 8-byte Folded Reload
@@ -296,10 +296,10 @@ define void @testi(ptr %a) {
; RV64D_NOPAIR-NEXT: .cfi_offset s3, -16
; RV64D_NOPAIR-NEXT: .cfi_offset s4, -24
; RV64D_NOPAIR-NEXT: .cfi_offset s5, -32
-; RV64D_NOPAIR-NEXT: ld s2, 8(a0)
; RV64D_NOPAIR-NEXT: ld s3, 0(a0)
-; RV64D_NOPAIR-NEXT: ld s4, 24(a0)
+; RV64D_NOPAIR-NEXT: ld s2, 8(a0)
; RV64D_NOPAIR-NEXT: ld s5, 16(a0)
+; RV64D_NOPAIR-NEXT: ld s4, 24(a0)
; RV64D_NOPAIR-NEXT: #APP
; RV64D_NOPAIR-NEXT: #NO_APP
; RV64D_NOPAIR-NEXT: ld s2, 24(sp) # 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/RISCV/rvv/vsetvl-cross-inline-asm.ll b/llvm/test/CodeGen/RISCV/rvv/vsetvl-cross-inline-asm.ll
index e5ba9a0a9bc5d2..21caa1a59465b5 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vsetvl-cross-inline-asm.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vsetvl-cross-inline-asm.ll
@@ -7,8 +7,8 @@ define void @foo(<vscale x 8 x half> %0) {
; CHECK-NEXT: vsetvli a0, zero, e32, m1, ta, ma
; CHECK-NEXT: vmv.v.i v10, 0
; CHECK-NEXT: lui a0, 1
-; CHECK-NEXT: addi a0, a0, -1096
; CHECK-NEXT: vmv.v.i v11, 0
+; CHECK-NEXT: addi a0, a0, -1096
; CHECK-NEXT: vsetvli zero, a0, e8, m1, ta, ma
; CHECK-NEXT: #APP
; CHECK-NEXT: vfmadd.vv v11, v10, v10
diff --git a/llvm/test/CodeGen/RISCV/tlsdesc-clobber-vector.ll b/llvm/test/CodeGen/RISCV/tlsdesc-clobber-vector.ll
index 4359af03b4a026..c689b7c5fbab7e 100644
--- a/llvm/test/CodeGen/RISCV/tlsdesc-clobber-vector.ll
+++ b/llvm/test/CodeGen/RISCV/tlsdesc-clobber-vector.ll
@@ -97,9 +97,6 @@ define void @test_vector_reg_clobber() nounwind {
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v1, (a0) # vscale x 8-byte Folded Reload
-; RV64-NEXT: #APP
-; RV64-NEXT: # use v1
-; RV64-NEXT: #NO_APP
; RV64-NEXT: csrr a0, vlenb
; RV64-NEXT: slli a0, a0, 1
; RV64-NEXT: mv a1, a0
@@ -108,52 +105,55 @@ define void @test_vector_reg_clobber() nounwind {
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v2, (a0) # vscale x 8-byte Folded Reload
-; RV64-NEXT: #APP
-; RV64-NEXT: # use v2
-; RV64-NEXT: #NO_APP
; RV64-NEXT: csrr a0, vlenb
; RV64-NEXT: slli a1, a0, 2
; RV64-NEXT: add a0, a1, a0
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v3, (a0) # vscale x 8-byte Folded Reload
-; RV64-NEXT: #APP
-; RV64-NEXT: # use v3
-; RV64-NEXT: #NO_APP
; RV64-NEXT: csrr a0, vlenb
; RV64-NEXT: slli a0, a0, 2
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v4, (a0) # vscale x 8-byte Folded Reload
-; RV64-NEXT: #APP
-; RV64-NEXT: # use v4
-; RV64-NEXT: #NO_APP
; RV64-NEXT: csrr a0, vlenb
; RV64-NEXT: slli a1, a0, 1
; RV64-NEXT: add a0, a1, a0
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v5, (a0) # vscale x 8-byte Folded Reload
-; RV64-NEXT: #APP
-; RV64-NEXT: # use v5
-; RV64-NEXT: #NO_APP
; RV64-NEXT: csrr a0, vlenb
; RV64-NEXT: slli a0, a0, 1
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v6, (a0) # vscale x 8-byte Folded Reload
-; RV64-NEXT: #APP
-; RV64-NEXT: # use v6
-; RV64-NEXT: #NO_APP
; RV64-NEXT: csrr a0, vlenb
; RV64-NEXT: add a0, sp, a0
; RV64-NEXT: addi a0, a0, 16
; RV64-NEXT: vl1r.v v7, (a0) # vscale x 8-byte Folded Reload
+; RV64-NEXT: addi a0, sp, 16
+; RV64-NEXT: vl1r.v v8, (a0) # vscale x 8-byte Folded Reload
+; RV64-NEXT: #APP
+; RV64-NEXT: # use v1
+; RV64-NEXT: #NO_APP
+; RV64-NEXT: #APP
+; RV64-NEXT: # use v2
+; RV64-NEXT: #NO_APP
+; RV64-NEXT: #APP
+; RV64-NEXT: # use v3
+; RV64-NEXT: #NO_APP
+; RV64-NEXT: #APP
+; RV64-NEXT: # use v4
+; RV64-NEXT: #NO_APP
+; RV64-NEXT: #APP
+; RV64-NEXT: # use v5
+; RV64-NEXT: #NO_APP
+; RV64-NEXT: #APP
+; RV64-NEXT: # use v6
+; RV64-NEXT: #NO_APP
; RV64-NEXT: #APP
; RV64-NEXT: # use v7
; RV64-NEXT: #NO_APP
-; RV64-NEXT: addi a0, sp, 16
-; RV64-NEXT: vl1r.v v8, (a0) # vscale x 8-byte Folded Reload
; RV64-NEXT: #APP
; RV64-NEXT: # use v8
; RV64-NEXT: #NO_APP
@@ -249,9 +249,6 @@ define void @test_vector_reg_clobber() nounwind {
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v1, (a0) # vscale x 8-byte Folded Reload
-; RV32-NEXT: #APP
-; RV32-NEXT: # use v1
-; RV32-NEXT: #NO_APP
; RV32-NEXT: csrr a0, vlenb
; RV32-NEXT: slli a0, a0, 1
; RV32-NEXT: mv a1, a0
@@ -260,52 +257,55 @@ define void @test_vector_reg_clobber() nounwind {
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v2, (a0) # vscale x 8-byte Folded Reload
-; RV32-NEXT: #APP
-; RV32-NEXT: # use v2
-; RV32-NEXT: #NO_APP
; RV32-NEXT: csrr a0, vlenb
; RV32-NEXT: slli a1, a0, 2
; RV32-NEXT: add a0, a1, a0
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v3, (a0) # vscale x 8-byte Folded Reload
-; RV32-NEXT: #APP
-; RV32-NEXT: # use v3
-; RV32-NEXT: #NO_APP
; RV32-NEXT: csrr a0, vlenb
; RV32-NEXT: slli a0, a0, 2
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v4, (a0) # vscale x 8-byte Folded Reload
-; RV32-NEXT: #APP
-; RV32-NEXT: # use v4
-; RV32-NEXT: #NO_APP
; RV32-NEXT: csrr a0, vlenb
; RV32-NEXT: slli a1, a0, 1
; RV32-NEXT: add a0, a1, a0
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v5, (a0) # vscale x 8-byte Folded Reload
-; RV32-NEXT: #APP
-; RV32-NEXT: # use v5
-; RV32-NEXT: #NO_APP
; RV32-NEXT: csrr a0, vlenb
; RV32-NEXT: slli a0, a0, 1
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v6, (a0) # vscale x 8-byte Folded Reload
-; RV32-NEXT: #APP
-; RV32-NEXT: # use v6
-; RV32-NEXT: #NO_APP
; RV32-NEXT: csrr a0, vlenb
; RV32-NEXT: add a0, sp, a0
; RV32-NEXT: addi a0, a0, 16
; RV32-NEXT: vl1r.v v7, (a0) # vscale x 8-byte Folded Reload
+; RV32-NEXT: addi a0, sp, 16
+; RV32-NEXT: vl1r.v v8, (a0) # vscale x 8-byte Folded Reload
+; RV32-NEXT: #APP
+; RV32-NEXT: # use v1
+; RV32-NEXT: #NO_APP
+; RV32-NEXT: #APP
+; RV32-NEXT: # use v2
+; RV32-NEXT: #NO_APP
+; RV32-NEXT: #APP
+; RV32-NEXT: # use v3
+; RV32-NEXT: #NO_APP
+; RV32-NEXT: #APP
+; RV32-NEXT: # use v4
+; RV32-NEXT: #NO_APP
+; RV32-NEXT: #APP
+; RV32-NEXT: # use v5
+; RV32-NEXT: #NO_APP
+; RV32-NEXT: #APP
+; RV32-NEXT: # use v6
+; RV32-NEXT: #NO_APP
; RV32-NEXT: #APP
; RV32-NEXT: # use v7
; RV32-NEXT: #NO_APP
-; RV32-NEXT: addi a0, sp, 16
-; RV32-NEXT: vl1r.v v8, (a0) # vscale x 8-byte Folded Reload
; RV32-NEXT: #APP
; RV32-NEXT: # use v8
; RV32-NEXT: #NO_APP
diff --git a/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting-explicit-regs.ll b/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting-explicit-regs.ll
index cf4dbbff8bec0b..e8aa0b1b5e6deb 100644
--- a/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting-explicit-regs.ll
+++ b/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting-explicit-regs.ll
@@ -49,8 +49,8 @@ define void @__int128_and_f(ptr noalias nocapture writeonly sret(i128) align 8 %
; Z15-LABEL: __int128_and_f:
; Z15: # %bb.0: # %entry
; Z15-NEXT: vl %v0, 0(%r3), 3
-; Z15-NEXT: vlr %v4, %v0
; Z15-NEXT: vrepg %v6, %v0, 1
+; Z15-NEXT: vlr %v4, %v0
; Z15-NEXT: #APP
; Z15-NEXT: #NO_APP
; Z15-NEXT: vmrhg %v0, %v4, %v6
@@ -300,8 +300,8 @@ entry:
define <4 x i32> @vec128_and_f(<4 x i32> %cc_dep1) {
; CHECK-LABEL: vec128_and_f:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vlr %v1, %v24
; CHECK-NEXT: vrepg %v3, %v24, 1
+; CHECK-NEXT: vlr %v1, %v24
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmrhg %v24, %v1, %v3
diff --git a/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting.ll b/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting.ll
index b23b40e0f0e907..61a828df25fac4 100644
--- a/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting.ll
+++ b/llvm/test/CodeGen/SystemZ/inline-asm-fp-int-casting.ll
@@ -301,8 +301,8 @@ entry:
define <4 x i32> @vec128_and_f(<4 x i32> %cc_dep1) {
; CHECK-LABEL: vec128_and_f:
; CHECK: # %bb.0: # %entry
-; CHECK-NEXT: vlr %v0, %v24
; CHECK-NEXT: vrepg %v2, %v24, 1
+; CHECK-NEXT: vlr %v0, %v24
; CHECK-NEXT: #APP
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: vmrhg %v24, %v0, %v2
diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/inlineasm.ll b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/inlineasm.ll
index 01411c896243df..1cf3f94be4dcca 100644
--- a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/inlineasm.ll
+++ b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/inlineasm.ll
@@ -11,9 +11,9 @@ define i32 @test(ptr nocapture readonly %x, ptr nocapture readonly %y, i32 %n) {
; CHECK-NEXT: blt .LBB0_2
; CHECK-NEXT: .LBB0_1: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: ldrh r3, [r1], #2
-; CHECK-NEXT: subs r2, #1
; CHECK-NEXT: ldrh r12, [r0], #2
+; CHECK-NEXT: subs r2, #1
+; CHECK-NEXT: ldrh r3, [r1], #2
; CHECK-NEXT: @APP
; CHECK-NEXT: add r3, r12
; CHECK-NEXT: @NO_APP
@@ -57,9 +57,9 @@ define i32 @testlr(ptr nocapture readonly %x, ptr nocapture readonly %y, i32 %n)
; CHECK-NEXT: movs r3, #0
; CHECK-NEXT: .LBB1_2: @ %for.body
; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT: ldrh r4, [r1], #2
-; CHECK-NEXT: subs r2, #1
; CHECK-NEXT: ldrh r12, [r0], #2
+; CHECK-NEXT: subs r2, #1
+; CHECK-NEXT: ldrh r4, [r1], #2
; CHECK-NEXT: @APP
; CHECK-NEXT: add r4, r12
; CHECK-NEXT: @NO_APP
diff --git a/llvm/test/CodeGen/Thumb2/arm_canberra_distance_f32.ll b/llvm/test/CodeGen/Thumb2/arm_canberra_distance_f32.ll
index c47a0bf6c00bd5..2bc85b2bfeefa8 100644
--- a/llvm/test/CodeGen/Thumb2/arm_canberra_distance_f32.ll
+++ b/llvm/test/CodeGen/Thumb2/arm_canberra_distance_f32.ll
@@ -24,10 +24,10 @@ define nofpclass(nan inf) float @arm_canberra_distance_f32(ptr noundef readonly
; CHECK-NEXT: @ %bb.2: @ %if.then
; CHECK-NEXT: @ in Loop: Header=BB0_1 Depth=1
; CHECK-NEXT: @APP
-; CHECK-NEXT: vabs.f32 s6, s4
+; CHECK-NEXT: vabs.f32 s8, s2
; CHECK-NEXT: @NO_APP
; CHECK-NEXT: @APP
-; CHECK-NEXT: vabs.f32 s8, s2
+; CHECK-NEXT: vabs.f32 s6, s4
; CHECK-NEXT: @NO_APP
; CHECK-NEXT: vsub.f32 s2, s4, s2
; CHECK-NEXT: vadd.f32 s6, s8, s6
diff --git a/llvm/test/CodeGen/X86/inlineasm-physreg-latency.mir b/llvm/test/CodeGen/X86/inlineasm-physreg-latency.mir
new file mode 100644
index 00000000000000..2aed3887e0412e
--- /dev/null
+++ b/llvm/test/CodeGen/X86/inlineasm-physreg-latency.mir
@@ -0,0 +1,173 @@
+# RUN: llc -mtriple=x86_64 -mcpu=skylake -run-pass=machine-scheduler -verify-misched -debug-only=machine-scheduler %s -filetype=null 2>&1 | FileCheck %s
+# RUN: llc -mtriple=x86_64 -mcpu=skylake -enable-post-misched -run-pass=postmisched -verify-misched -debug-only=machine-scheduler %s -filetype=null 2>&1 | FileCheck %s
+# REQUIRES: asserts
+
+# Check physical-register input latency for inline asm, including explicit
+# and fixed-register asm outputs, real implicit definitions, and trailing
+# register bookkeeping operands. Native uses keep their existing policy.
+
+---
+name: native_load_use
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: native_load_use:%bb.0
+ ; CHECK: SU(0): {{.*}}MOV64rm
+ ; CHECK: SU(1): Data Latency=5 Reg=$rax
+ bb.0:
+ liveins: $rdi
+ $rax = MOV64rm $rdi, 1, $noreg, 0, $noreg :: (load (s64))
+ $rcx = MOV64rr $rax
+ RET 0
+...
+
+---
+name: explicit_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: explicit_input:%bb.0
+ ; CHECK: SU(0): {{.*}}MOV64rm
+ ; CHECK: SU(1): Data Latency=5 Reg=$rax
+ bb.0:
+ liveins: $rdi
+ $rax = MOV64rm $rdi, 1, $noreg, 0, $noreg :: (load (s64))
+ INLINEASM &"# use $0", attdialect, reguse:GR64, $rax
+ RET 0
+...
+
+---
+name: tied_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: tied_input:%bb.0
+ ; CHECK: SU(0): {{.*}}MOV64rm
+ ; CHECK: SU(1): Data Latency=5 Reg=$rax
+ bb.0:
+ liveins: $rdi
+ $rax = MOV64rm $rdi, 1, $noreg, 0, $noreg :: (load (s64))
+ INLINEASM &"", attdialect, regdef:GR64, def $rax, reguse tiedto:$0, $rax(tied-def 3)
+ RET 0
+...
+
+---
+name: implicit_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: implicit_input:%bb.0
+ ; CHECK: SU(0): {{.*}}MOV64rm
+ ; CHECK: SU(1): Data Latency=0 Reg=$rax
+ bb.0:
+ liveins: $rdi
+ $rax = MOV64rm $rdi, 1, $noreg, 0, $noreg :: (load (s64))
+ INLINEASM &"", attdialect, implicit $rax
+ RET 0
+...
+
+---
+name: implicit_pseudo_def
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: implicit_pseudo_def:%bb.0
+ ; CHECK: SU(0): {{.*}}MOV32ri
+ ; CHECK: SU(1): Data Latency=0 Reg=$ecx
+ bb.0:
+ $eax = MOV32ri 1, implicit-def $ecx
+ INLINEASM &"# use $0", attdialect, reguse:GR32, $ecx
+ RET 0
+...
+
+---
+name: real_implicit_def
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: real_implicit_def:%bb.0
+ ; CHECK: SU(0): {{.*}}ADD32ri8
+ ; CHECK: SU(1): Data Latency=1 Reg=$eflags
+ bb.0:
+ liveins: $eax
+ $eax = ADD32ri8 $eax, 1, implicit-def $eflags
+ INLINEASM &"# use flags", attdialect, reguse:CCR, $eflags
+ RET 0
+...
+
+---
+name: asm_output_to_asm_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: asm_output_to_asm_input:%bb.0
+ ; CHECK: SU(0): {{.*}}INLINEASM
+ ; CHECK: SU(1): Data Latency=1 Reg=$rax
+ bb.0:
+ INLINEASM &"# def $0", attdialect, regdef:GR64, def $rax
+ INLINEASM &"# use $0", attdialect, reguse:GR64, $rax
+ RET 0
+...
+
+---
+name: asm_output_to_native_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: asm_output_to_native_input:%bb.0
+ ; CHECK: SU(0): {{.*}}INLINEASM
+ ; CHECK: SU(1): Data Latency=0 Reg=$rax
+ bb.0:
+ INLINEASM &"# def $0", attdialect, regdef:GR64, def $rax
+ $rcx = MOV64rr $rax
+ RET 0
+...
+
+# Fixed-register output constraints are emitted with implicit-def on the real
+# output operand, inside its regdef group. These dependencies have the same
+# latency as asm_output_to_asm_input above.
+---
+name: asm_fixed_output_to_asm_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: asm_fixed_output_to_asm_input:%bb.0
+ ; CHECK: SU(0): {{.*}}INLINEASM
+ ; CHECK: SU(1): Data Latency=1 Reg=$rax
+ bb.0:
+ ; Output constraint: "={rax}".
+ INLINEASM &"mov $$1, $0", sideeffect attdialect, regdef, implicit-def $rax
+ INLINEASM &"# use $0", sideeffect attdialect, reguse, $rax
+ RET 0
+...
+
+---
+name: asm_fixed_earlyclobber_output_to_asm_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: asm_fixed_earlyclobber_output_to_asm_input:%bb.0
+ ; CHECK: SU(0): {{.*}}INLINEASM
+ ; CHECK: SU(1): Data Latency=1 Reg=$rax
+ bb.0:
+ ; Output constraint: "=&{rax}".
+ INLINEASM &"mov $$1, $0", sideeffect attdialect, regdef-ec, implicit-def early-clobber $rax
+ INLINEASM &"# use $0", sideeffect attdialect, reguse, $rax
+ RET 0
+...
+
+---
+name: asm_implicit_pseudo_def
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: asm_implicit_pseudo_def:%bb.0
+ ; CHECK: SU(0): {{.*}}INLINEASM
+ ; CHECK: SU(1): Data Latency=0 Reg=$rax
+ bb.0:
+ INLINEASM &"", attdialect, implicit-def $rax
+ INLINEASM &"# use $0", attdialect, reguse:GR64, $rax
+ RET 0
+...
+
+---
+name: asm_output_to_implicit_input
+tracksRegLiveness: true
+body: |
+ ; CHECK-LABEL: asm_output_to_implicit_input:%bb.0
+ ; CHECK: SU(0): {{.*}}INLINEASM
+ ; CHECK: SU(1): Data Latency=0 Reg=$rax
+ bb.0:
+ INLINEASM &"# def $0", attdialect, regdef:GR64, def $rax
+ INLINEASM &"", attdialect, implicit $rax
+ RET 0
+...
>From 8f275192fc497d6b609389151cc50187843fd0bc Mon Sep 17 00:00:00 2001
From: David Green <david.green at arm.com>
Date: Sat, 21 Nov 2026 14:09:53 +0000
Subject: [PATCH 24/35] [ARM] Promote bf16 libcalls via f32 (#244879)
This add promotion rules for bf16 operations, the list of which is taken
from the current fp16 operations.
---
llvm/lib/Target/ARM/ARMISelLowering.cpp | 16 +
llvm/test/CodeGen/ARM/bf16-instructions.ll | 552 ++++++++++++++++++---
2 files changed, 490 insertions(+), 78 deletions(-)
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index f3d6881dfc2b08..0207fadf757051 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -588,6 +588,22 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM_,
ISD::FFLOOR, ISD::FNEARBYINT, ISD::FRINT, ISD::FROUND,
ISD::FROUNDEVEN, ISD::FTRUNC})
setOperationAction(Op, MVT::bf16, Promote);
+
+ for (auto Op : {ISD::FREM, ISD::FPOW, ISD::FPOWI,
+ ISD::FCOS, ISD::FSIN, ISD::FSINCOS,
+ ISD::FSINCOSPI, ISD::FMODF, ISD::FACOS,
+ ISD::FASIN, ISD::FATAN, ISD::FATAN2,
+ ISD::FCOSH, ISD::FSINH, ISD::FTANH,
+ ISD::FTAN, ISD::FEXP, ISD::FEXP2,
+ ISD::FEXP10, ISD::FLOG, ISD::FLOG2,
+ ISD::FLOG10, ISD::STRICT_FREM, ISD::STRICT_FPOW,
+ ISD::STRICT_FPOWI, ISD::STRICT_FCOS, ISD::STRICT_FSIN,
+ ISD::STRICT_FACOS, ISD::STRICT_FASIN, ISD::STRICT_FATAN,
+ ISD::STRICT_FATAN2, ISD::STRICT_FCOSH, ISD::STRICT_FSINH,
+ ISD::STRICT_FTANH, ISD::STRICT_FEXP, ISD::STRICT_FEXP2,
+ ISD::STRICT_FLOG, ISD::STRICT_FLOG2, ISD::STRICT_FLOG10,
+ ISD::STRICT_FTAN})
+ setOperationAction(Op, MVT::bf16, Promote);
} else {
setOperationAction(ISD::BF16_TO_FP, MVT::f32, Expand);
setOperationAction(ISD::BF16_TO_FP, MVT::f64, Expand);
diff --git a/llvm/test/CodeGen/ARM/bf16-instructions.ll b/llvm/test/CodeGen/ARM/bf16-instructions.ll
index cadc16b4740916..a18a1c720466d1 100644
--- a/llvm/test/CodeGen/ARM/bf16-instructions.ll
+++ b/llvm/test/CodeGen/ARM/bf16-instructions.ll
@@ -275,10 +275,37 @@ define bfloat @test_fdiv(bfloat %a, bfloat %b) #0 {
ret bfloat %r
}
-;define bfloat @test_frem(bfloat %a, bfloat %b) #0 {
-; %r = frem bfloat %a, %b
-; ret bfloat %r
-;}
+define bfloat @test_frem(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_frem:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: lsl r1, r1, #16
+; CHECK-CVT-NEXT: vmov s1, r0
+; CHECK-CVT-NEXT: vmov s0, r1
+; CHECK-CVT-NEXT: bl fmodf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_frem:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s1
+; CHECK-BF16-NEXT: vmov r1, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: lsl r1, r1, #16
+; CHECK-BF16-NEXT: vmov s1, r0
+; CHECK-BF16-NEXT: vmov s0, r1
+; CHECK-BF16-NEXT: bl fmodf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = frem bfloat %a, %b
+ ret bfloat %r
+}
define bfloat @test_load(ptr %a) #0 {
; CHECK-CVT-LABEL: test_load:
@@ -555,7 +582,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-CVT-NEXT: mov r4, r0
; CHECK-CVT-NEXT: ldrh r0, [r0]
; CHECK-CVT-NEXT: vmov s18, r0
-; CHECK-CVT-NEXT: .LBB11_1: @ %loop
+; CHECK-CVT-NEXT: .LBB12_1: @ %loop
; CHECK-CVT-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-CVT-NEXT: ldrh r0, [r4]
; CHECK-CVT-NEXT: vmov.f32 s16, s18
@@ -563,7 +590,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-CVT-NEXT: mov r0, r4
; CHECK-CVT-NEXT: bl test_dummy
; CHECK-CVT-NEXT: tst r0, #1
-; CHECK-CVT-NEXT: bne .LBB11_1
+; CHECK-CVT-NEXT: bne .LBB12_1
; CHECK-CVT-NEXT: @ %bb.2: @ %return
; CHECK-CVT-NEXT: vmov.f32 s0, s16
; CHECK-CVT-NEXT: vpop {d8, d9}
@@ -578,7 +605,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-NOFP16-NEXT: mov r4, r0
; CHECK-NOFP16-NEXT: ldrh r0, [r0]
; CHECK-NOFP16-NEXT: vmov s18, r0
-; CHECK-NOFP16-NEXT: .LBB11_1: @ %loop
+; CHECK-NOFP16-NEXT: .LBB12_1: @ %loop
; CHECK-NOFP16-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-NOFP16-NEXT: ldrh r0, [r4]
; CHECK-NOFP16-NEXT: vmov.f32 s16, s18
@@ -586,7 +613,7 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-NOFP16-NEXT: mov r0, r4
; CHECK-NOFP16-NEXT: bl test_dummy
; CHECK-NOFP16-NEXT: tst r0, #1
-; CHECK-NOFP16-NEXT: bne .LBB11_1
+; CHECK-NOFP16-NEXT: bne .LBB12_1
; CHECK-NOFP16-NEXT: @ %bb.2: @ %return
; CHECK-NOFP16-NEXT: vmov.f32 s0, s16
; CHECK-NOFP16-NEXT: vpop {d8, d9}
@@ -600,14 +627,14 @@ define bfloat @test_phi(ptr %p1) #0 {
; CHECK-FP16-NEXT: vpush {d8, d9}
; CHECK-FP16-NEXT: vldr.16 s18, [r0]
; CHECK-FP16-NEXT: mov r4, r0
-; CHECK-FP16-NEXT: .LBB11_1: @ %loop
+; CHECK-FP16-NEXT: .LBB12_1: @ %loop
; CHECK-FP16-NEXT: @ =>This Inner Loop Header: Depth=1
; CHECK-FP16-NEXT: mov r0, r4
; CHECK-FP16-NEXT: vmov.f32 s16, s18
; CHECK-FP16-NEXT: vldr.16 s18, [r4]
; CHECK-FP16-NEXT: bl test_dummy
; CHECK-FP16-NEXT: tst r0, #1
-; CHECK-FP16-NEXT: bne .LBB11_1
+; CHECK-FP16-NEXT: bne .LBB12_1
; CHECK-FP16-NEXT: @ %bb.2: @ %return
; CHECK-FP16-NEXT: vmov.f32 s0, s16
; CHECK-FP16-NEXT: vpop {d8, d9}
@@ -837,90 +864,459 @@ define bfloat @test_bitcast_f16tobfloat(half %a) #0 {
; ret bfloat %r
;}
-;define bfloat @test_powi(bfloat %a, i32 %b) #0 {
-; %r = call bfloat @llvm.powi.bf16.i32(bfloat %a, i32 %b)
-; ret bfloat %r
-;}
+define bfloat @test_powi(bfloat %a, i32 %b) #0 {
+; CHECK-CVT-LABEL: test_powi:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r1, r1, #16
+; CHECK-CVT-NEXT: vmov s0, r1
+; CHECK-CVT-NEXT: bl __powisf2
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_powi:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r1, s0
+; CHECK-BF16-NEXT: lsl r1, r1, #16
+; CHECK-BF16-NEXT: vmov s0, r1
+; CHECK-BF16-NEXT: bl __powisf2
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.powi.bf16.i32(bfloat %a, i32 %b)
+ ret bfloat %r
+}
-;define bfloat @test_sin(bfloat %a) #0 {
-; %r = call bfloat @llvm.sin.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_sin(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_sin:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl sinf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_sin:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl sinf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.sin.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_cos(bfloat %a) #0 {
-; %r = call bfloat @llvm.cos.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_cos(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_cos:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl cosf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_cos:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl cosf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.cos.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_tan(bfloat %a) #0 {
-; %r = call bfloat @llvm.tan.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_tan(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_tan:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl tanf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_tan:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl tanf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.tan.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_acos(bfloat %a) #0 {
-; %r = call bfloat @llvm.acos.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_acos(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_acos:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl acosf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_acos:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl acosf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.acos.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_asin(bfloat %a) #0 {
-; %r = call bfloat @llvm.asin.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_asin(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_asin:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl asinf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_asin:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl asinf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.asin.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_atan(bfloat %a) #0 {
-; %r = call bfloat @llvm.atan.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_atan(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_atan:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl atanf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_atan:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl atanf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.atan.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_atan2(bfloat %a, bfloat %b) #0 {
-; %r = call bfloat @llvm.atan2.bf16(bfloat %a, bfloat %b)
-; ret bfloat %r
-;}
+define bfloat @test_atan2(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_atan2:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: lsl r1, r1, #16
+; CHECK-CVT-NEXT: vmov s1, r0
+; CHECK-CVT-NEXT: vmov s0, r1
+; CHECK-CVT-NEXT: bl atan2f
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_atan2:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s1
+; CHECK-BF16-NEXT: vmov r1, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: lsl r1, r1, #16
+; CHECK-BF16-NEXT: vmov s1, r0
+; CHECK-BF16-NEXT: vmov s0, r1
+; CHECK-BF16-NEXT: bl atan2f
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.atan2.bf16(bfloat %a, bfloat %b)
+ ret bfloat %r
+}
-;define bfloat @test_cosh(bfloat %a) #0 {
-; %r = call bfloat @llvm.cosh.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_cosh(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_cosh:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl coshf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_cosh:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl coshf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.cosh.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_sinh(bfloat %a) #0 {
-; %r = call bfloat @llvm.sinh.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_sinh(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_sinh:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl sinhf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_sinh:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl sinhf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.sinh.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_tanh(bfloat %a) #0 {
-; %r = call bfloat @llvm.tanh.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_tanh(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_tanh:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl tanhf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_tanh:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl tanhf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.tanh.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_pow(bfloat %a, bfloat %b) #0 {
-; %r = call bfloat @llvm.pow.bf16(bfloat %a, bfloat %b)
-; ret bfloat %r
-;}
+define bfloat @test_pow(bfloat %a, bfloat %b) #0 {
+; CHECK-CVT-LABEL: test_pow:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s1
+; CHECK-CVT-NEXT: vmov r1, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: lsl r1, r1, #16
+; CHECK-CVT-NEXT: vmov s1, r0
+; CHECK-CVT-NEXT: vmov s0, r1
+; CHECK-CVT-NEXT: bl powf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_pow:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s1
+; CHECK-BF16-NEXT: vmov r1, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: lsl r1, r1, #16
+; CHECK-BF16-NEXT: vmov s1, r0
+; CHECK-BF16-NEXT: vmov s0, r1
+; CHECK-BF16-NEXT: bl powf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.pow.bf16(bfloat %a, bfloat %b)
+ ret bfloat %r
+}
-;define bfloat @test_exp(bfloat %a) #0 {
-; %r = call bfloat @llvm.exp.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_exp(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_exp:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl expf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_exp:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl expf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.exp.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_exp2(bfloat %a) #0 {
-; %r = call bfloat @llvm.exp2.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_exp2(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_exp2:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl exp2f
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_exp2:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl exp2f
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.exp2.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_log(bfloat %a) #0 {
-; %r = call bfloat @llvm.log.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_log(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_log:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl logf
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_log:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl logf
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.log.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_log10(bfloat %a) #0 {
-; %r = call bfloat @llvm.log10.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_log10(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_log10:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl log10f
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_log10:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl log10f
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.log10.bf16(bfloat %a)
+ ret bfloat %r
+}
-;define bfloat @test_log2(bfloat %a) #0 {
-; %r = call bfloat @llvm.log2.bf16(bfloat %a)
-; ret bfloat %r
-;}
+define bfloat @test_log2(bfloat %a) #0 {
+; CHECK-CVT-LABEL: test_log2:
+; CHECK-CVT: @ %bb.0:
+; CHECK-CVT-NEXT: .save {r11, lr}
+; CHECK-CVT-NEXT: push {r11, lr}
+; CHECK-CVT-NEXT: vmov r0, s0
+; CHECK-CVT-NEXT: lsl r0, r0, #16
+; CHECK-CVT-NEXT: vmov s0, r0
+; CHECK-CVT-NEXT: bl log2f
+; CHECK-CVT-NEXT: bl __truncsfbf2
+; CHECK-CVT-NEXT: pop {r11, pc}
+;
+; CHECK-BF16-LABEL: test_log2:
+; CHECK-BF16: @ %bb.0:
+; CHECK-BF16-NEXT: .save {r11, lr}
+; CHECK-BF16-NEXT: push {r11, lr}
+; CHECK-BF16-NEXT: vmov r0, s0
+; CHECK-BF16-NEXT: lsl r0, r0, #16
+; CHECK-BF16-NEXT: vmov s0, r0
+; CHECK-BF16-NEXT: bl log2f
+; CHECK-BF16-NEXT: vcvtb.bf16.f32 s0, s0
+; CHECK-BF16-NEXT: pop {r11, pc}
+ %r = call bfloat @llvm.log2.bf16(bfloat %a)
+ ret bfloat %r
+}
;define bfloat @test_fma(bfloat %a, bfloat %b, bfloat %c) #0 {
; %r = call bfloat @llvm.fma.bf16(bfloat %a, bfloat %b, bfloat %c)
>From 9ce6459837dfaf6e72efca54e09bb9915dfdfb4a Mon Sep 17 00:00:00 2001
From: "Michael G. Kazakov" <mike.kazakov at gmail.com>
Date: Sat, 21 Nov 2026 14:16:37 +0000
Subject: [PATCH 25/35] [libc++][pstl] Add a serial implementation of
std::rotate (#244574)
This PR adds a serial implementation of `std::rotate` to PSTL.
Currently there's no intent to provide a parallel version of this
algorithm, but later it can be added as well.
Part of #99938.
---
libcxx/include/__algorithm/pstl.h | 12 +
libcxx/include/__pstl/backend_fwd.h | 6 +
libcxx/include/__pstl/backends/serial.h | 10 +
.../pstl.iterator-requirements.verify.cpp | 4 +
.../alg.rotate/pstl.rotate.pass.cpp | 341 ++++++++++++++++++
.../pstl.exception_handling.pass.cpp | 9 +
6 files changed, 382 insertions(+)
create mode 100644 libcxx/test/std/algorithms/alg.modifying.operations/alg.rotate/pstl.rotate.pass.cpp
diff --git a/libcxx/include/__algorithm/pstl.h b/libcxx/include/__algorithm/pstl.h
index 96cb210b2d6fc8..0486d2a3044258 100644
--- a/libcxx/include/__algorithm/pstl.h
+++ b/libcxx/include/__algorithm/pstl.h
@@ -1009,6 +1009,18 @@ void replace_copy(_ExecutionPolicy&& __policy,
__new_value);
}
+template <class _ExecutionPolicy,
+ class _ForwardIterator,
+ class _RawPolicy = __remove_cvref_t<_ExecutionPolicy>,
+ enable_if_t<is_execution_policy_v<_RawPolicy>, int> = 0>
+_ForwardIterator
+rotate(_ExecutionPolicy&& __policy, _ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last) {
+ _LIBCPP_REQUIRE_CPP17_FORWARD_ITERATOR(_ForwardIterator, "rotate requires ForwardIterators");
+ using _Implementation = __pstl::__dispatch<__pstl::__rotate, __pstl::__current_configuration, _RawPolicy>;
+ return __pstl::__handle_exception<_Implementation>(
+ std::forward<_ExecutionPolicy>(__policy), std::move(__first), std::move(__middle), std::move(__last));
+}
+
template <class _ExecutionPolicy,
class _ForwardIterator,
class _ForwardOutIterator,
diff --git a/libcxx/include/__pstl/backend_fwd.h b/libcxx/include/__pstl/backend_fwd.h
index bf153706453173..6053b3617a2dcd 100644
--- a/libcxx/include/__pstl/backend_fwd.h
+++ b/libcxx/include/__pstl/backend_fwd.h
@@ -415,6 +415,12 @@ struct __reverse_copy;
// operator()(_Policy&&, _BidirectionalIterator __first, _BidirectionalIterator __last,
// _ForwardIterator __result) const noexcept;
+template <class _Backend, class _ExecutionPolicy>
+struct __rotate;
+// template <class _Policy, class _ForwardIterator>
+// optional<_ForwardIterator>
+// operator()(_Policy&&, _ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last) const noexcept;
+
template <class _Backend, class _ExecutionPolicy>
struct __rotate_copy;
// template <class _Policy, class _ForwardIterator, class _ForwardOutIterator>
diff --git a/libcxx/include/__pstl/backends/serial.h b/libcxx/include/__pstl/backends/serial.h
index d5205777d95cfe..f932f5bdd81cb5 100644
--- a/libcxx/include/__pstl/backends/serial.h
+++ b/libcxx/include/__pstl/backends/serial.h
@@ -23,6 +23,7 @@
#include <__algorithm/partition.h>
#include <__algorithm/partition_copy.h>
#include <__algorithm/reverse.h>
+#include <__algorithm/rotate.h>
#include <__algorithm/search.h>
#include <__algorithm/search_n.h>
#include <__algorithm/set_difference.h>
@@ -179,6 +180,15 @@ struct __reverse<__serial_backend_tag, _ExecutionPolicy> {
}
};
+template <class _ExecutionPolicy>
+struct __rotate<__serial_backend_tag, _ExecutionPolicy> {
+ template <class _Policy, class _ForwardIterator>
+ optional<_ForwardIterator>
+ operator()(_Policy&&, _ForwardIterator __first, _ForwardIterator __middle, _ForwardIterator __last) const noexcept {
+ return std::rotate(std::move(__first), std::move(__middle), std::move(__last));
+ }
+};
+
template <class _ExecutionPolicy>
struct __search<__serial_backend_tag, _ExecutionPolicy> {
template <class _Policy, class _ForwardIterator1, class _ForwardIterator2, class _BinaryPredicate>
diff --git a/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp b/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
index 038f8f61a803f4..b44cafd4564949 100644
--- a/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
+++ b/libcxx/test/libcxx/algorithms/pstl.iterator-requirements.verify.cpp
@@ -315,6 +315,10 @@ void f(non_forward_iterator non_fwd,
pol, it, it, non_output, val, val); // expected-error@*:* {{static assertion failed: replace_copy}}
}
+ {
+ (void)std::rotate(pol, non_fwd, non_fwd, non_fwd); // expected-error@*:* {{static assertion failed: rotate}}
+ }
+
{
(void)std::rotate_copy(
pol, non_fwd, non_fwd, non_fwd, out); // expected-error@*:* {{static assertion failed: rotate_copy}}
diff --git a/libcxx/test/std/algorithms/alg.modifying.operations/alg.rotate/pstl.rotate.pass.cpp b/libcxx/test/std/algorithms/alg.modifying.operations/alg.rotate/pstl.rotate.pass.cpp
new file mode 100644
index 00000000000000..098d1d1c6ee3ee
--- /dev/null
+++ b/libcxx/test/std/algorithms/alg.modifying.operations/alg.rotate/pstl.rotate.pass.cpp
@@ -0,0 +1,341 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// REQUIRES: std-at-least-c++17
+
+// UNSUPPORTED: libcpp-has-no-incomplete-pstl
+
+// <algorithm>
+
+// template<class ExecutionPolicy, class ForwardIterator>
+// ForwardIterator
+// rotate(ExecutionPolicy&& exec,
+// ForwardIterator first, ForwardIterator middle, ForwardIterator last);
+
+#include <algorithm>
+#include <cassert>
+#include <functional>
+#include <iterator>
+#include <limits>
+#include <memory>
+#include <numeric>
+#include <type_traits>
+
+#include "test_execution_policies.h"
+#include "test_iterators.h"
+#include "test_macros.h"
+#include "type_algorithms.h"
+
+EXECUTION_POLICY_SFINAE_TEST(rotate);
+
+static_assert(sfinae_test_rotate<int, int*, int*, int*>);
+static_assert(!sfinae_test_rotate<std::execution::parallel_policy, int*, int*, int*>);
+
+template <class Iter>
+struct TestInt {
+ template <class ExecutionPolicy>
+ void operator()(ExecutionPolicy&& policy) {
+ { // Check the return type
+ int a[] = {0};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ static_assert(std::is_same_v<decltype(r), Iter>);
+ }
+ { // Check empty
+ int a[] = {0};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::begin(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ }
+ { // Check single element, rotate at first
+ int a[] = {0};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(a[0] == 0);
+ }
+ { // Check single element, rotate at last
+ int a[] = {0};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ }
+ { // Check two elements, rotate at first
+ int a[] = {0, 1};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ }
+ { // Check two elements, rotate at second
+ int a[] = {0, 1};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 1), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 1));
+ assert(a[0] == 1);
+ assert(a[1] == 0);
+ }
+ { // Check two elements, rotate at last
+ int a[] = {0, 1};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ }
+ { // Check three elements, rotate at first
+ int a[] = {0, 1, 2};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ }
+ { // Check three elements, rotate at second
+ int a[] = {0, 1, 2};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 1), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 2));
+ assert(a[0] == 1);
+ assert(a[1] == 2);
+ assert(a[2] == 0);
+ }
+ { // Check three elements, rotate at third
+ int a[] = {0, 1, 2};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 2), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 1));
+ assert(a[0] == 2);
+ assert(a[1] == 0);
+ assert(a[2] == 1);
+ }
+ { // Check three elements, rotate at last
+ int a[] = {0, 1, 2};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ }
+ { // Check four elements, rotate at first
+ int a[] = {0, 1, 2, 3};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ assert(a[3] == 3);
+ }
+ { // Check four elements, rotate at second
+ int a[] = {0, 1, 2, 3};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 1), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 3));
+ assert(a[0] == 1);
+ assert(a[1] == 2);
+ assert(a[2] == 3);
+ assert(a[3] == 0);
+ }
+ { // Check four elements, rotate at third
+ int a[] = {0, 1, 2, 3};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 2), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 2));
+ assert(a[0] == 2);
+ assert(a[1] == 3);
+ assert(a[2] == 0);
+ assert(a[3] == 1);
+ }
+ { // Check four elements, rotate at fourth
+ int a[] = {0, 1, 2, 3};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 3), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 1));
+ assert(a[0] == 3);
+ assert(a[1] == 0);
+ assert(a[2] == 1);
+ assert(a[3] == 2);
+ }
+ { // Check four elements, rotate at last
+ int a[] = {0, 1, 2, 3};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ assert(a[3] == 3);
+ }
+ { // Check five elements, rotate at first
+ int a[] = {0, 1, 2, 3, 4};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ assert(a[3] == 3);
+ assert(a[4] == 4);
+ }
+ { // Check five elements, rotate at second
+ int a[] = {0, 1, 2, 3, 4};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 1), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 4));
+ assert(a[0] == 1);
+ assert(a[1] == 2);
+ assert(a[2] == 3);
+ assert(a[3] == 4);
+ assert(a[4] == 0);
+ }
+ { // Check five elements, rotate at third
+ int a[] = {0, 1, 2, 3, 4};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 2), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 3));
+ assert(a[0] == 2);
+ assert(a[1] == 3);
+ assert(a[2] == 4);
+ assert(a[3] == 0);
+ assert(a[4] == 1);
+ }
+ { // Check five elements, rotate at fourth
+ int a[] = {0, 1, 2, 3, 4};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 3), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 2));
+ assert(a[0] == 3);
+ assert(a[1] == 4);
+ assert(a[2] == 0);
+ assert(a[3] == 1);
+ assert(a[4] == 2);
+ }
+ { // Check five elements, rotate at fifth
+ int a[] = {0, 1, 2, 3, 4};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 4), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 1));
+ assert(a[0] == 4);
+ assert(a[1] == 0);
+ assert(a[2] == 1);
+ assert(a[3] == 2);
+ assert(a[4] == 3);
+ }
+ { // Check five elements, rotate at last
+ int a[] = {0, 1, 2, 3, 4};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ assert(a[3] == 3);
+ assert(a[4] == 4);
+ }
+ { // Check six elements, rotate at first
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ assert(a[3] == 3);
+ assert(a[4] == 4);
+ assert(a[5] == 5);
+ }
+ { // Check six elements, rotate at second
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 1), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 5));
+ assert(a[0] == 1);
+ assert(a[1] == 2);
+ assert(a[2] == 3);
+ assert(a[3] == 4);
+ assert(a[4] == 5);
+ assert(a[5] == 0);
+ }
+ { // Check six elements, rotate at third
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 2), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 4));
+ assert(a[0] == 2);
+ assert(a[1] == 3);
+ assert(a[2] == 4);
+ assert(a[3] == 5);
+ assert(a[4] == 0);
+ assert(a[5] == 1);
+ }
+ { // Check six elements, rotate at fourth
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 3), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 3));
+ assert(a[0] == 3);
+ assert(a[1] == 4);
+ assert(a[2] == 5);
+ assert(a[3] == 0);
+ assert(a[4] == 1);
+ assert(a[5] == 2);
+ }
+ { // Check six elements, rotate at fifth
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 4), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 2));
+ assert(a[0] == 4);
+ assert(a[1] == 5);
+ assert(a[2] == 0);
+ assert(a[3] == 1);
+ assert(a[4] == 2);
+ assert(a[5] == 3);
+ }
+ { // Check six elements, rotate at sixth
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 5), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 1));
+ assert(a[0] == 5);
+ assert(a[1] == 0);
+ assert(a[2] == 1);
+ assert(a[3] == 2);
+ assert(a[4] == 3);
+ assert(a[5] == 4);
+ }
+ { // Check six elements, rotate at sixth
+ int a[] = {0, 1, 2, 3, 4, 5};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(a[0] == 0);
+ assert(a[1] == 1);
+ assert(a[2] == 2);
+ assert(a[3] == 3);
+ assert(a[4] == 4);
+ assert(a[5] == 5);
+ }
+ }
+};
+
+template <class Iter>
+struct TestNonCopyable {
+ template <class ExecutionPolicy>
+ void operator()(ExecutionPolicy&& policy) {
+ { // Check the return type
+ std::unique_ptr<int> a[] = {std::make_unique<int>(0)};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ static_assert(std::is_same_v<decltype(r), Iter>);
+ }
+ { // Check two elements, rotate at first
+ std::unique_ptr<int> a[] = {std::make_unique<int>(0), std::make_unique<int>(1)};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a)), Iter(std::end(a)));
+ assert(r == Iter(std::end(a)));
+ assert(*a[0] == 0);
+ assert(*a[1] == 1);
+ }
+ { // Check two elements, rotate at second
+ std::unique_ptr<int> a[] = {std::make_unique<int>(0), std::make_unique<int>(1)};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::begin(a) + 1), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a) + 1));
+ assert(*a[0] == 1);
+ assert(*a[1] == 0);
+ }
+ { // Check two elements, rotate at last
+ std::unique_ptr<int> a[] = {std::make_unique<int>(0), std::make_unique<int>(1)};
+ auto r = std::rotate(policy, Iter(std::begin(a)), Iter(std::end(a)), Iter(std::end(a)));
+ assert(r == Iter(std::begin(a)));
+ assert(*a[0] == 0);
+ assert(*a[1] == 1);
+ }
+ }
+};
+
+int main(int, char**) {
+ types::for_each(types::forward_iterator_list<int*>{}, TestIteratorWithPolicies<TestInt>{});
+ types::for_each(types::forward_iterator_list<std::unique_ptr<int>*>{}, TestIteratorWithPolicies<TestNonCopyable>{});
+ return 0;
+}
diff --git a/libcxx/test/std/algorithms/pstl.exception_handling.pass.cpp b/libcxx/test/std/algorithms/pstl.exception_handling.pass.cpp
index 69aaa1a6363719..2ed7848cb092a0 100644
--- a/libcxx/test/std/algorithms/pstl.exception_handling.pass.cpp
+++ b/libcxx/test/std/algorithms/pstl.exception_handling.pass.cpp
@@ -491,6 +491,15 @@ int main(int, char**) {
});
}
+ {
+ auto mid1 = util::throw_on_move_iterator(std::begin(a) + 2, tokens[5].active() ? 1 : -1);
+
+ // rotate(first, mid, last)
+ assert_non_throwing([=, &policy] {
+ (void)std::rotate(policy, std::move(first1), std::move(mid1), std::move(last1));
+ });
+ }
+
{
auto mid1 = util::throw_on_move_iterator(std::begin(a) + 2, tokens[5].active() ? 1 : -1);
>From 0c41652846ac0771141157712a31b6ff29e78a0c Mon Sep 17 00:00:00 2001
From: Dhairyashil R G <dhairyashil25 at gmail.com>
Date: Sat, 21 Nov 2026 19:47:41 +0530
Subject: [PATCH 26/35] [mlir][affine] Mask transfers in SuperVectorize instead
of using in_bounds (#243425)
SuperVectorize sets `in_bounds` when the memref dim is a multiple of the
vector size, without looking at the index. With vectors of 8 on a
`memref<64xf32>`, a read of `%src[%i + 1]` reads one element past the
end, and a write to `%dst[%i + 4]` writes four past the end.
SuperVectorize now masks each transfer by `dim - index`. For that,
`createReadOrMaskedRead` now builds its mask like
`createWriteOrMaskedWrite`, from the indices and the permutation map.
SuperVectorize is the only caller that passes them.
For downstream pipelines: the masks are `vector.mask` regions, so
lowering to LLVM needs `-lower-vector-mask`. `-eliminate-vector-masks
-canonicalize` removes the masks that are provably all-true and leaves
those transfers with `in_bounds = [true]`. Main left some of them at the
default `false`, e.g. a read of `memref<100xf32>` in a loop `0 to 64`.
Transfer unrolling handles `vector.mask` since #239554.
Tests: `vectorize_masked_read_write.mlir` is new, the existing tests now
expect the masks, and `vectorize_transpose_2d.mlir` adds the transposing
write promised in #233244.
#221268 fixed the same bug with `in_bounds`. Tim096 will close it once
this lands.
Part of #227821.
Assisted-by: Claude
---
.../mlir/Dialect/Vector/Utils/VectorUtils.h | 20 ++--
.../Affine/Transforms/SuperVectorize.cpp | 7 +-
mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp | 101 +++++++++++-------
.../Affine/SuperVectorize/transform_op.mlir | 16 +--
.../Affine/SuperVectorize/vector_utils.mlir | 8 +-
.../Affine/SuperVectorize/vectorize_1d.mlir | 42 ++++----
.../Affine/SuperVectorize/vectorize_2d.mlir | 20 ++--
.../SuperVectorize/vectorize_2d_inbounds.mlir | 36 +++++--
.../Affine/SuperVectorize/vectorize_3d.mlir | 2 +-
.../vectorize_affine_apply.mlir | 10 +-
.../vectorize_masked_read_write.mlir | 55 ++++++++++
.../vectorize_outer_loop_2d.mlir | 2 +-
.../vectorize_outer_loop_transpose_2d.mlir | 8 +-
.../SuperVectorize/vectorize_reduction.mlir | 53 ++++-----
.../vectorize_transpose_2d.mlir | 27 ++++-
.../Linalg/vectorization/insert-slice.mlir | 9 +-
16 files changed, 262 insertions(+), 154 deletions(-)
create mode 100644 mlir/test/Dialect/Affine/SuperVectorize/vectorize_masked_read_write.mlir
diff --git a/mlir/include/mlir/Dialect/Vector/Utils/VectorUtils.h b/mlir/include/mlir/Dialect/Vector/Utils/VectorUtils.h
index 050296449eb23e..6ceb4ea7fa4bb0 100644
--- a/mlir/include/mlir/Dialect/Vector/Utils/VectorUtils.h
+++ b/mlir/include/mlir/Dialect/Vector/Utils/VectorUtils.h
@@ -242,16 +242,16 @@ bool isLinearizableVector(VectorType type);
/// Creates a TransferReadOp from `source`.
///
-/// If the shape of vector to read differs from the shape of the value being
-/// read, masking is used to avoid out-of-bounds accesses. Set
-/// `useInBoundsInsteadOfMasking` to `true` to use the "in_bounds" attribute
-/// instead of explicit masks.
-///
-/// When `permutationMap` is provided the in_bounds attribute is inferred from
-/// it: dimension i is in-bounds when the map result is an AffineDimExpr
-/// pointing to a static memref dimension divisible by the vector size, or an
-/// AffineConstantExpr (broadcast). Custom`indices` must also be supplied in
-/// that case; if `indices` is empty, all offsets default to 0.
+/// The read is masked unless it is statically in bounds. The mask has a size
+/// of `dim - index` for each `source` dim that `permutationMap` selects (all
+/// dims without a map). Set `useInBoundsInsteadOfMasking` to `true` to use the
+/// "in_bounds" attribute instead of a mask.
+///
+/// In that case, when `permutationMap` is provided the in_bounds attribute is
+/// inferred from it: dimension i is in-bounds when the map result is an
+/// AffineDimExpr pointing to a static memref dimension divisible by the vector
+/// size, or an AffineConstantExpr (broadcast). Custom `indices` must also be
+/// supplied in that case; if `indices` is empty, all offsets default to 0.
Value createReadOrMaskedRead(OpBuilder &builder, Location loc, Value source,
const VectorType &vecToReadTy,
std::optional<Value> padValue = std::nullopt,
diff --git a/mlir/lib/Dialect/Affine/Transforms/SuperVectorize.cpp b/mlir/lib/Dialect/Affine/Transforms/SuperVectorize.cpp
index e2c0a3c6ab4fa9..863aaf50804e44 100644
--- a/mlir/lib/Dialect/Affine/Transforms/SuperVectorize.cpp
+++ b/mlir/lib/Dialect/Affine/Transforms/SuperVectorize.cpp
@@ -126,6 +126,9 @@ using namespace vector;
/// In the future, these operations will expose a contract to constrain the
/// search on vectorization patterns and sizes.
///
+/// Each transfer is emitted under a `vector.mask` that enables the lanes inside
+/// the memref. The examples in this file leave the mask out.
+///
/// Occurrence of super-vectorization in the compiler flow:
/// =======================================================
/// This is an active area of investigation. We start with 2 remarks to position
@@ -1275,7 +1278,7 @@ static Operation *vectorizeAffineLoad(AffineLoadOp loadOp,
Value transferVal = createReadOrMaskedRead(
state.builder, loadOp.getLoc(), loadOp.getMemRef(), vectorType,
- /*padValue=*/std::nullopt, /*useInBoundsInsteadOfMasking=*/true, indices,
+ /*padValue=*/std::nullopt, /*useInBoundsInsteadOfMasking=*/false, indices,
permutationMap);
Operation *transfer = transferVal.getDefiningOp();
@@ -1334,7 +1337,7 @@ static Operation *vectorizeAffineStore(AffineStoreOp storeOp,
Operation *transfer = createWriteOrMaskedWrite(
state.builder, storeOp.getLoc(), vectorValue, storeOp.getMemRef(),
SmallVector<Value>(indices.begin(), indices.end()),
- /*useInBoundsInsteadOfMasking=*/true, permutationMap);
+ /*useInBoundsInsteadOfMasking=*/false, permutationMap);
LLVM_DEBUG(dbgs() << "\n[early-vect]+++++ vectorized store: " << *transfer);
// Register replacement for future uses in the scope.
diff --git a/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp b/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp
index 51a32a6771fe78..60ae63c82f2624 100644
--- a/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp
+++ b/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp
@@ -444,6 +444,56 @@ static SmallVector<bool> computeInBoundsFromPermutationMap(
return inBounds;
}
+/// Returns the mask for a transfer on `base` at `indices` with permutation map
+/// `map`, or null if the transfer needs none.
+static Value createTransferMask(OpBuilder &builder, Location loc,
+ VectorType vecType, Value base,
+ ArrayRef<Value> indices, AffineMap map) {
+ // The map has a dim per base dim but a result per vector dim, and the vector
+ // can have fewer dims than the base (e.g. rank-reducing `insert_slice`). The
+ // mask has a dim for each base dim that the map selects, in increasing order.
+ SmallVector<int64_t> maskedDims;
+ for (AffineExpr expr : map.getResults())
+ if (auto dimExpr = dyn_cast<AffineDimExpr>(expr))
+ maskedDims.push_back(dimExpr.getPosition());
+ // A read that only broadcasts has nothing to mask.
+ if (maskedDims.empty())
+ return Value();
+ // A transposing map selects them out of order.
+ llvm::sort(maskedDims);
+
+ ArrayRef<int64_t> baseShape = cast<ShapedType>(base.getType()).getShape();
+
+ // For each masked dim, collect the mask size (`dim - index`), index and base
+ // size. A mask size known statically creates no op, so none is left unused
+ // when the mask folds away.
+ SmallVector<OpFoldResult> maskSizes;
+ SmallVector<Value> maskedIndices;
+ SmallVector<int64_t> maskedShape;
+ for (int64_t dim : maskedDims) {
+ std::optional<int64_t> cstIndex = getConstantIntValue(indices[dim]);
+ if (ShapedType::isStatic(baseShape[dim]) && cstIndex) {
+ maskSizes.push_back(builder.getIndexAttr(baseShape[dim] - *cstIndex));
+ } else {
+ OpFoldResult size = isa<MemRefType>(base.getType())
+ ? memref::getMixedSize(builder, loc, base, dim)
+ : tensor::getMixedSize(builder, loc, base, dim);
+ maskSizes.push_back(builder.createOrFold<arith::SubIOp>(
+ loc, getValueOrCreateConstantIndexOp(builder, loc, size),
+ indices[dim]));
+ }
+ maskedIndices.push_back(indices[dim]);
+ maskedShape.push_back(baseShape[dim]);
+ }
+
+ VectorType maskType = vector::inferTransferOpMaskType(vecType, map);
+ if (isMaskTriviallyFoldable(maskSizes, maskedIndices, maskedShape,
+ maskType.getShape()))
+ return Value();
+
+ return builder.createOrFold<vector::CreateMaskOp>(loc, maskType, maskSizes);
+}
+
Value vector::createReadOrMaskedRead(OpBuilder &builder, Location loc,
Value source,
const VectorType &vecToReadTy,
@@ -512,18 +562,15 @@ Value vector::createReadOrMaskedRead(OpBuilder &builder, Location loc,
if (useInBoundsInsteadOfMasking)
return transferReadOp;
- SmallVector<OpFoldResult> mixedSourceDims =
- isa<MemRefType>(source.getType())
- ? memref::getMixedSizes(builder, loc, source)
- : tensor::getMixedSizes(builder, loc, source);
-
- if (isMaskTriviallyFoldable(mixedSourceDims, indices, sourceShape,
- vecToReadShape))
+ // A null map means an identity: every source dim is read.
+ AffineMap map = permutationMap
+ ? permutationMap
+ : builder.getMultiDimIdentityMap(vecToReadRank);
+ Value mask =
+ createTransferMask(builder, loc, vecToReadTy, source, indices, map);
+ if (!mask)
return transferReadOp;
- auto maskType = vecToReadTy.cloneWith(/*shape=*/{}, builder.getI1Type());
- Value mask =
- vector::CreateMaskOp::create(builder, loc, maskType, mixedSourceDims);
return mlir::vector::maskOperation(builder, transferReadOp, mask)
->getResult(0);
}
@@ -600,39 +647,11 @@ Operation *vector::createWriteOrMaskedWrite(OpBuilder &builder, Location loc,
return write;
// Compute the mask and mask the write Op.
- VectorType writeMaskType = inferTransferOpMaskType(vecToStoreType, map);
-
- SmallVector<OpFoldResult> destSizes =
- isa<MemRefType>(dest.getType())
- ? memref::getMixedSizes(builder, loc, dest)
- : tensor::getMixedSizes(builder, loc, dest);
-
- // The map has a dim per dest dim but a result per vector dim, and the vector
- // can have fewer dims than the dest (e.g. rank-reducing `insert_slice`). The
- // mask has a dim for each dest dim that the map selects, in increasing order.
- // For each, collect the mask size (`dim - index`), write index and dest size.
- SmallVector<int64_t> maskedDestDims;
- for (AffineExpr expr : map.getResults())
- maskedDestDims.push_back(cast<AffineDimExpr>(expr).getPosition());
- // A transposing map selects them out of order.
- llvm::sort(maskedDestDims);
- SmallVector<OpFoldResult> maskSizes;
- SmallVector<Value> maskedDestIndices;
- SmallVector<int64_t> maskedDestShape;
- for (int64_t dim : maskedDestDims) {
- Value size = getValueOrCreateConstantIndexOp(builder, loc, destSizes[dim]);
- maskSizes.push_back(
- builder.createOrFold<arith::SubIOp>(loc, size, writeIndices[dim]));
- maskedDestIndices.push_back(writeIndices[dim]);
- maskedDestShape.push_back(destShape[dim]);
- }
-
- if (isMaskTriviallyFoldable(maskSizes, maskedDestIndices, maskedDestShape,
- writeMaskType.getShape()))
+ Value maskForWrite =
+ createTransferMask(builder, loc, vecToStoreType, dest, writeIndices, map);
+ if (!maskForWrite)
return write;
- Value maskForWrite =
- builder.createOrFold<vector::CreateMaskOp>(loc, writeMaskType, maskSizes);
return mlir::vector::maskOperation(builder, write, maskForWrite);
}
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/transform_op.mlir b/mlir/test/Dialect/Affine/SuperVectorize/transform_op.mlir
index 5734362ee855fe..0514f86e6a018d 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/transform_op.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/transform_op.mlir
@@ -9,7 +9,7 @@ func.func @vec1d_3(%A : memref<?x?xf32>, %B : memref<?x?x?xf32>) {
%N = memref.dim %A, %c1 : memref<?x?xf32>
%P = memref.dim %B, %c2 : memref<?x?x?xf32>
-// CHECK: %{{.*}} = vector.transfer_read {{.*}} : memref<?x?xf32>, vector<128xf32>
+// CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read {{.*}} : memref<?x?xf32>, vector<128xf32> }
affine.for %i8 = 0 to %M { // vectorized
affine.for %i9 = 0 to %N {
%a9 = affine.load %A[%i9, %i8 + %i9] : memref<?x?xf32>
@@ -35,17 +35,17 @@ func.func @vectorize_matmul(%arg0: memref<?x?xf32>, %arg1: memref<?x?xf32>, %arg
%M = memref.dim %arg0, %c0 : memref<?x?xf32>
%K = memref.dim %arg0, %c1 : memref<?x?xf32>
%N = memref.dim %arg2, %c1 : memref<?x?xf32>
- // CHECK: vector.transfer_write {{.*}} : vector<4x8xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write {{.*}} : vector<4x8xf32>, memref<?x?xf32> }
affine.for %i0 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%M) {
affine.for %i1 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%N) {
%cst = arith.constant 0.000000e+00 : f32
affine.store %cst, %arg2[%i0, %i1] : memref<?x?xf32>
}
}
- // CHECK: %{{.*}} = vector.transfer_read {{.*}} : memref<?x?xf32>, vector<4x8xf32>
- // CHECK: %{{.*}} = vector.transfer_read {{.*}} : memref<?x?xf32>, vector<4x8xf32>
- // CHECK: %{{.*}} = vector.transfer_read {{.*}} : memref<?x?xf32>, vector<4x8xf32>
- // CHECK: vector.transfer_write %{{.*}} : vector<4x8xf32>, memref<?x?xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read {{.*}} : memref<?x?xf32>, vector<4x8xf32> }
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read {{.*}} : memref<?x?xf32>, vector<4x8xf32> }
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read {{.*}} : memref<?x?xf32>, vector<4x8xf32> }
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %{{.*}} : vector<4x8xf32>, memref<?x?xf32> }
affine.for %i2 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%M) {
affine.for %i3 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%N) {
affine.for %i4 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%K) {
@@ -78,7 +78,7 @@ func.func @vec3d(%A : memref<?x?x?xf32>) {
%0 = memref.dim %A, %c0 : memref<?x?x?xf32>
%1 = memref.dim %A, %c1 : memref<?x?x?xf32>
%2 = memref.dim %A, %c2 : memref<?x?x?xf32>
- // CHECK: %{{.*}} = vector.transfer_read {{.*}} : memref<?x?x?xf32>, vector<32x64x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read {{.*}} : memref<?x?x?xf32>, vector<32x64x256xf32> }
affine.for %t0 = 0 to %0 {
affine.for %t1 = 0 to %0 {
affine.for %i0 = 0 to %0 {
@@ -117,7 +117,7 @@ func.func @vecdim_reduction_minf(%in: memref<256x512xf32>, %out: memref<256xf32>
}
// CHECK-LABEL: @vecdim_reduction_minf
-// CHECK: %{{.*}} = vector.transfer_read {{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read {{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %{{.*}} = vector.reduction <minimumf>, {{.*}} : vector<128xf32> into f32
module attributes {transform.with_named_sequence} {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vector_utils.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vector_utils.mlir
index 71d8314bb60d8b..c11abe136471a1 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vector_utils.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vector_utils.mlir
@@ -70,10 +70,10 @@ func.func @transfer_rank_mismatch_no_crash(%arg0: memref<82x97xf32>) {
}
// VECNEST: affine.for %{{.*}} = 0 to 20 step 4 {
-// VECNEST: vector.transfer_read
+// VECNEST: vector.mask %{{.*}} { vector.transfer_read
// VECNEST-NEXT: affine.for %{{.*}} = 0 to 30 {
-// VECNEST: vector.transfer_read
-// VECNEST-NEXT: vector.transfer_write %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}] {in_bounds = [true], permutation_map = #{{.*}}}
+// VECNEST: vector.mask %{{.*}} { vector.transfer_read
+// VECNEST: vector.mask %{{.*}} { vector.transfer_write %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}] {in_bounds = [true], permutation_map = #{{.*}}}
// VECNEST-NEXT: }
-// VECNEST-NEXT: vector.transfer_write
+// VECNEST: vector.mask %{{.*}} { vector.transfer_write
// VECNEST: }
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_1d.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_1d.mlir
index f99190be6365dd..88d8c641f16174 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_1d.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_1d.mlir
@@ -48,7 +48,7 @@ func.func @vec1d_2(%A : memref<?x?xf32>, %B : memref<?x?x?xf32>) {
// CHECK:for [[IV3:%[a-zA-Z0-9]+]] = 0 to [[ARG_M]] step 128
// CHECK-NEXT: %[[CST:.*]] = ub.poison : f32
-// CHECK-NEXT: {{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}], %[[CST]] : memref<?x?xf32>, vector<128xf32>
+// CHECK: {{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}], %[[CST]] {in_bounds = [true]} : memref<?x?xf32>, vector<128xf32> }
affine.for %i3 = 0 to %M { // vectorized
%a3 = affine.load %A[%c0, %i3] : memref<?x?xf32>
}
@@ -77,7 +77,7 @@ func.func @vec1d_3(%A : memref<?x?xf32>, %B : memref<?x?x?xf32>) {
// CHECK-NEXT: %[[APP9_0:[0-9a-zA-Z_]+]] = affine.apply {{.*}}([[IV9]], [[IV8]])
// CHECK-NEXT: %[[APP9_1:[0-9a-zA-Z_]+]] = affine.apply {{.*}}([[IV9]], [[IV8]])
// CHECK-NEXT: %[[CST:.*]] = ub.poison : f32
-// CHECK-NEXT: {{.*}} = vector.transfer_read %{{.*}}[%[[APP9_0]], %[[APP9_1]]], %[[CST]] : memref<?x?xf32>, vector<128xf32>
+// CHECK: {{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%[[APP9_0]], %[[APP9_1]]], %[[CST]] {in_bounds = [true]} : memref<?x?xf32>, vector<128xf32> }
affine.for %i8 = 0 to %M { // vectorized
affine.for %i9 = 0 to %N {
%a9 = affine.load %A[%i9, %i8 + %i9] : memref<?x?xf32>
@@ -98,7 +98,7 @@ func.func @vector_add_2d(%M : index, %N : index) -> f32 {
affine.for %i0 = 0 to %M {
affine.for %i1 = 0 to %N {
// CHECK: %[[C1:.*]] = arith.constant dense<1.000000e+00> : vector<128xf32>
- // CHECK: vector.transfer_write %[[C1]], {{.*}} : vector<128xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %[[C1]], {{.*}} : vector<128xf32>, memref<?x?xf32> }
// non-scoped %f1
affine.store %f1, %A[%i0, %i1] : memref<?x?xf32, 0>
}
@@ -106,7 +106,7 @@ func.func @vector_add_2d(%M : index, %N : index) -> f32 {
affine.for %i2 = 0 to %M {
affine.for %i3 = 0 to %N {
// CHECK: %[[C3:.*]] = arith.constant dense<2.000000e+00> : vector<128xf32>
- // CHECK: vector.transfer_write %[[C3]], {{.*}} : vector<128xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %[[C3]], {{.*}} : vector<128xf32>, memref<?x?xf32> }
// non-scoped %f2
affine.store %f2, %B[%i2, %i3] : memref<?x?xf32, 0>
}
@@ -115,13 +115,13 @@ func.func @vector_add_2d(%M : index, %N : index) -> f32 {
affine.for %i5 = 0 to %N {
// CHECK: %[[SPLAT2:.*]] = arith.constant dense<2.000000e+00> : vector<128xf32>
// CHECK: %[[SPLAT1:.*]] = arith.constant dense<1.000000e+00> : vector<128xf32>
- // CHECK: %[[A5:.*]] = vector.transfer_read %{{.*}}[{{.*}}], %{{[a-zA-Z0-9_]*}} : memref<?x?xf32>, vector<128xf32>
- // CHECK: %[[B5:.*]] = vector.transfer_read %{{.*}}[{{.*}}], %{{[a-zA-Z0-9_]*}} : memref<?x?xf32>, vector<128xf32>
+ // CHECK: %[[A5:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[{{.*}}], %{{[a-zA-Z0-9_]*}} {in_bounds = [true]} : memref<?x?xf32>, vector<128xf32> }
+ // CHECK: %[[B5:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[{{.*}}], %{{[a-zA-Z0-9_]*}} {in_bounds = [true]} : memref<?x?xf32>, vector<128xf32> }
// CHECK: %[[S5:.*]] = arith.addf %[[A5]], %[[B5]] fastmath<nnan> {test.marker} : vector<128xf32>
// CHECK: %[[S6:.*]] = arith.addf %[[S5]], %[[SPLAT1]] : vector<128xf32>
// CHECK: %[[S7:.*]] = arith.addf %[[S5]], %[[SPLAT2]] : vector<128xf32>
// CHECK: %[[S8:.*]] = arith.addf %[[S7]], %[[S6]] : vector<128xf32>
- // CHECK: vector.transfer_write %[[S8]], {{.*}} : vector<128xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %[[S8]], {{.*}} : vector<128xf32>, memref<?x?xf32> }
%a5 = affine.load %A[%i4, %i5] : memref<?x?xf32, 0>
%b5 = affine.load %B[%i4, %i5] : memref<?x?xf32, 0>
%s5 = arith.addf %a5, %b5 fastmath<nnan> {test.marker} : f32
@@ -150,11 +150,11 @@ func.func @vec_constant_with_two_users(%M : index, %N : index) -> (f32, f32) {
affine.for %i0 = 0 to %M { // vectorized
// CHECK: %[[C1:.*]] = arith.constant dense<1.000000e+00> : vector<128xf32>
// CHECK-NEXT: affine.for
- // CHECK-NEXT: vector.transfer_write %[[C1]], {{.*}} : vector<128xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %[[C1]], {{.*}} : vector<128xf32>, memref<?x?xf32> }
affine.for %i1 = 0 to %N {
affine.store %f1, %A[%i1, %i0] : memref<?x?xf32, 0>
}
- // CHECK: vector.transfer_write %[[C1]], {{.*}} : vector<128xf32>, memref<?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %[[C1]], {{.*}} : vector<128xf32>, memref<?xf32> }
affine.store %f1, %B[%i0] : memref<?xf32, 0>
}
%c12 = arith.constant 12 : index
@@ -171,7 +171,7 @@ func.func @vec_block_arg(%A : memref<32x512xi32>) {
// CHECK-NEXT: affine.for %[[IV1:[0-9a-zA-Z_]+]] = 0 to 32 {
// CHECK-NEXT: %[[BROADCAST:.*]] = vector.broadcast %[[IV1]] : index to vector<128xindex>
// CHECK-NEXT: %[[CAST:.*]] = arith.index_cast %[[BROADCAST]] : vector<128xindex> to vector<128xi32>
- // CHECK-NEXT: vector.transfer_write %[[CAST]], {{.*}}[%[[IV1]], %[[IV0]]] {in_bounds = [true]} : vector<128xi32>, memref<32x512xi32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write %[[CAST]], {{.*}}[%[[IV1]], %[[IV0]]] {in_bounds = [true]} : vector<128xi32>, memref<32x512xi32> }
affine.for %i = 0 to 512 { // vectorized
affine.for %j = 0 to 32 {
%idx = arith.index_cast %j : index to i32
@@ -197,7 +197,7 @@ func.func @vec_block_arg_2(%A : memref<?x512xindex>) {
// CHECK-NEXT: %[[BROADCAST2:.*]] = vector.broadcast %[[IV2]] : index to vector<128xindex>
// CHECK-NEXT: %[[INDEX1:.*]] = affine.apply #[[$map0]](%[[IV0]], %[[IV2]], %[[IV1]])
// CHECK-NEXT: %[[INDEX2:.*]] = affine.apply #[[$map1]](%[[IV0]], %[[IV2]], %[[IV1]])
- // CHECK: %[[LOAD:.*]] = vector.transfer_read %{{.*}}[%[[INDEX1]], %[[INDEX2]]], %{{.*}} : memref<?x512xindex>, vector<128xindex>
+ // CHECK: %[[LOAD:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%[[INDEX1]], %[[INDEX2]]], %{{.*}} : memref<?x512xindex>, vector<128xindex> }
// CHECK-NEXT: arith.muli %[[BROADCAST1]], %[[LOAD]] : vector<128xindex>
// CHECK-NEXT: arith.addi %{{.*}}, %[[BROADCAST2]] : vector<128xindex>
// CHECK: }
@@ -281,7 +281,7 @@ func.func @vec_rejected_3(%A : memref<?x?xf32>, %B : memref<?x?x?xf32>) {
// CHECK:for [[IV4:%[0-9a-zA-Z_]+]] = 0 to [[ARG_M]] step 128 {
// CHECK-NEXT: for [[IV5:%[0-9a-zA-Z_]*]] = 0 to [[ARG_N]] {
// CHECK-NEXT: %{{.*}} = ub.poison : f32
-// CHECK-NEXT: {{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}], %{{[a-zA-Z0-9_]*}} : memref<?x?xf32>, vector<128xf32>
+// CHECK: {{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}], %{{[a-zA-Z0-9_]*}} {in_bounds = [true]} : memref<?x?xf32>, vector<128xf32> }
affine.for %i4 = 0 to %M { // vectorized
affine.for %i5 = 0 to %N { // not vectorized, would vectorize with --test-fastest-varying=1
%a5 = affine.load %A[%i5, %i4] : memref<?x?xf32>
@@ -592,11 +592,11 @@ func.func @vec_non_vecdim_reduction(%in: memref<128x256xf32>, %out: memref<256xf
// CHECK: affine.for %{{.*}} = 0 to 256 step 128 {
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[final_red:.*]] = affine.for %{{.*}} = 0 to 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<128x256xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<128x256xf32>, vector<128xf32> }
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
// CHECK: }
-// CHECK: vector.transfer_write %[[final_red]], %{{.*}} : vector<128xf32>, memref<256xf32>
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[final_red]], %{{.*}} : vector<128xf32>, memref<256xf32> }
// CHECK: }
// -----
@@ -628,14 +628,14 @@ func.func @vec_non_vecdim_reductions(%in0: memref<128x256xf32>, %in1: memref<128
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[reds:.*]]:2 = affine.for %{{.*}} = 0 to 128
// CHECK-SAME: iter_args(%[[red_iter0:.*]] = %[[vzero]], %[[red_iter1:.*]] = %[[vone]]) -> (vector<128xf32>, vector<128xi32>) {
-// CHECK: %[[ld0:.*]] = vector.transfer_read %{{.*}} : memref<128x256xf32>, vector<128xf32>
+// CHECK: %[[ld0:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<128x256xf32>, vector<128xf32> }
// CHECK: %[[add:.*]] = arith.addf %[[red_iter0]], %[[ld0]] : vector<128xf32>
-// CHECK: %[[ld1:.*]] = vector.transfer_read %{{.*}} : memref<128x256xi32>, vector<128xi32>
+// CHECK: %[[ld1:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<128x256xi32>, vector<128xi32> }
// CHECK: %[[mul:.*]] = arith.muli %[[red_iter1]], %[[ld1]] : vector<128xi32>
// CHECK: affine.yield %[[add]], %[[mul]] : vector<128xf32>, vector<128xi32>
// CHECK: }
-// CHECK: vector.transfer_write %[[reds]]#0, %{{.*}} : vector<128xf32>, memref<256xf32>
-// CHECK: vector.transfer_write %[[reds]]#1, %{{.*}} : vector<128xi32>, memref<256xi32>
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[reds]]#0, %{{.*}} : vector<128xf32>, memref<256xf32> }
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[reds]]#1, %{{.*}} : vector<128xi32>, memref<256xi32> }
// CHECK: }
// -----
@@ -658,10 +658,10 @@ func.func @vec_no_vecdim_last_value(%in: memref<128x256xf32>, %out: memref<256xf
// CHECK: affine.for %{{.*}} = 0 to 256 step 128 {
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[last_val:.*]] = affine.for %{{.*}} = 0 to 128 iter_args(%[[last_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<128x256xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<128x256xf32>, vector<128xf32> }
// CHECK: affine.yield %[[ld]] : vector<128xf32>
// CHECK: }
-// CHECK: vector.transfer_write %[[last_val]], %{{.*}} : vector<128xf32>, memref<256xf32>
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[last_val]], %{{.*}} : vector<128xf32>, memref<256xf32> }
// CHECK: }
// -----
@@ -712,7 +712,7 @@ func.func @vec_non_scalar_type() {
// CHECK: affine.for %[[IV:.*]] = 0 to 8 step 128 {
// CHECK-DAG: %[[CST_VEC:.*]] = arith.constant dense<0> : vector<128xi32>
// CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
-// CHECK: vector.transfer_write %[[CST_VEC]], %{{.*}}[%[[C0]], %[[IV]]] : vector<128xi32>, memref<1x8xi32>
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[CST_VEC]], %{{.*}}[%[[C0]], %[[IV]]] {in_bounds = [true]} : vector<128xi32>, memref<1x8xi32> }
// CHECK: }
func.func @index_const_inside_loop(%mem: memref<1x8xi32>) {
affine.for %arg0 = 0 to 8 {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d.mlir
index eb5120a49e3d4b..6b491c139d30a7 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d.mlir
@@ -55,7 +55,7 @@ func.func @vector_add_2d(%M : index, %N : index) -> f32 {
affine.for %i0 = 0 to %M {
affine.for %i1 = 0 to %N {
// CHECK: [[C1:%.*]] = arith.constant dense<1.000000e+00> : vector<32x256xf32>
- // CHECK: vector.transfer_write [[C1]], {{.*}} : vector<32x256xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write [[C1]], {{.*}} : vector<32x256xf32>, memref<?x?xf32> }
// non-scoped %f1
affine.store %f1, %A[%i0, %i1] : memref<?x?xf32, 0>
}
@@ -63,7 +63,7 @@ func.func @vector_add_2d(%M : index, %N : index) -> f32 {
affine.for %i2 = 0 to %M {
affine.for %i3 = 0 to %N {
// CHECK: [[C3:%.*]] = arith.constant dense<2.000000e+00> : vector<32x256xf32>
- // CHECK: vector.transfer_write [[C3]], {{.*}} : vector<32x256xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write [[C3]], {{.*}} : vector<32x256xf32>, memref<?x?xf32> }
// non-scoped %f2
affine.store %f2, %B[%i2, %i3] : memref<?x?xf32, 0>
}
@@ -72,13 +72,13 @@ func.func @vector_add_2d(%M : index, %N : index) -> f32 {
affine.for %i5 = 0 to %N {
// CHECK: [[SPLAT2:%.*]] = arith.constant dense<2.000000e+00> : vector<32x256xf32>
// CHECK: [[SPLAT1:%.*]] = arith.constant dense<1.000000e+00> : vector<32x256xf32>
- // CHECK: [[A5:%.*]] = vector.transfer_read %{{.*}}[{{.*}}], %{{.*}} : memref<?x?xf32>, vector<32x256xf32>
- // CHECK: [[B5:%.*]] = vector.transfer_read %{{.*}}[{{.*}}], %{{.*}} : memref<?x?xf32>, vector<32x256xf32>
+ // CHECK: [[A5:%.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[{{.*}}], %{{.*}} : memref<?x?xf32>, vector<32x256xf32> }
+ // CHECK: [[B5:%.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[{{.*}}], %{{.*}} : memref<?x?xf32>, vector<32x256xf32> }
// CHECK: [[S5:%.*]] = arith.addf [[A5]], [[B5]] : vector<32x256xf32>
// CHECK: [[S6:%.*]] = arith.addf [[S5]], [[SPLAT1]] : vector<32x256xf32>
// CHECK: [[S7:%.*]] = arith.addf [[S5]], [[SPLAT2]] : vector<32x256xf32>
// CHECK: [[S8:%.*]] = arith.addf [[S7]], [[S6]] : vector<32x256xf32>
- // CHECK: vector.transfer_write [[S8]], {{.*}} : vector<32x256xf32>, memref<?x?xf32>
+ // CHECK: vector.mask %{{.*}} { vector.transfer_write [[S8]], {{.*}} : vector<32x256xf32>, memref<?x?xf32> }
//
%a5 = affine.load %A[%i4, %i5] : memref<?x?xf32, 0>
%b5 = affine.load %B[%i4, %i5] : memref<?x?xf32, 0>
@@ -113,7 +113,7 @@ func.func @vectorize_matmul(%arg0: memref<?x?xf32>, %arg1: memref<?x?xf32>, %arg
// VECT: {{.*}} #[[$map_id1]](%[[M]]) step 4 {
// VECT-NEXT: {{.*}} #[[$map_id1]](%[[N]]) step 8 {
// VECT: %[[VC0:.*]] = arith.constant dense<0.000000e+00> : vector<4x8xf32>
- // VECT-NEXT: vector.transfer_write %[[VC0]], %{{.*}}[%{{.*}}, %{{.*}}] : vector<4x8xf32>, memref<?x?xf32>
+ // VECT: vector.mask %{{.*}} { vector.transfer_write %[[VC0]], %{{.*}}[%{{.*}}, %{{.*}}] {in_bounds = [true, true]} : vector<4x8xf32>, memref<?x?xf32> }
affine.for %i0 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%M) {
affine.for %i1 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%N) {
%cst = arith.constant 0.000000e+00 : f32
@@ -123,12 +123,12 @@ func.func @vectorize_matmul(%arg0: memref<?x?xf32>, %arg1: memref<?x?xf32>, %arg
// VECT: affine.for %[[I2:.*]] = #[[$map_id1]](%[[C0]]) to #[[$map_id1]](%[[M]]) step 4 {
// VECT-NEXT: affine.for %[[I3:.*]] = #[[$map_id1]](%[[C0]]) to #[[$map_id1]](%[[N]]) step 8 {
// VECT-NEXT: affine.for %[[I4:.*]] = #[[$map_id1]](%[[C0]]) to #[[$map_id1]](%[[K]]) {
- // VECT: %[[A:.*]] = vector.transfer_read %{{.*}}[%[[I4]], %[[I3]]], %{{.*}} {in_bounds = [true, false], permutation_map = #[[$map_proj_d0d1_zerod1]]} : memref<?x?xf32>, vector<4x8xf32>
- // VECT: %[[B:.*]] = vector.transfer_read %{{.*}}[%[[I2]], %[[I4]]], %{{.*}} {in_bounds = [false, true], permutation_map = #[[$map_proj_d0d1_d0zero]]} : memref<?x?xf32>, vector<4x8xf32>
+ // VECT: %[[A:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%[[I4]], %[[I3]]], %{{.*}} {in_bounds = [true, true], permutation_map = #[[$map_proj_d0d1_zerod1]]} : memref<?x?xf32>, vector<4x8xf32> }
+ // VECT: %[[B:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%[[I2]], %[[I4]]], %{{.*}} {in_bounds = [true, true], permutation_map = #[[$map_proj_d0d1_d0zero]]} : memref<?x?xf32>, vector<4x8xf32> }
// VECT-NEXT: %[[C:.*]] = arith.mulf %[[B]], %[[A]] : vector<4x8xf32>
- // VECT: %[[D:.*]] = vector.transfer_read %{{.*}}[%[[I2]], %[[I3]]], %{{.*}} : memref<?x?xf32>, vector<4x8xf32>
+ // VECT: %[[D:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%[[I2]], %[[I3]]], %{{.*}} : memref<?x?xf32>, vector<4x8xf32> }
// VECT-NEXT: %[[E:.*]] = arith.addf %[[D]], %[[C]] : vector<4x8xf32>
- // VECT: vector.transfer_write %[[E]], %{{.*}}[%[[I2]], %[[I3]]] : vector<4x8xf32>, memref<?x?xf32>
+ // VECT: vector.mask %{{.*}} { vector.transfer_write %[[E]], %{{.*}}[%[[I2]], %[[I3]]] {in_bounds = [true, true]} : vector<4x8xf32>, memref<?x?xf32> }
affine.for %i2 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%M) {
affine.for %i3 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%N) {
affine.for %i4 = affine_map<(d0) -> (d0)>(%c0) to affine_map<(d0) -> (d0)>(%K) {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d_inbounds.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d_inbounds.mlir
index 36bdb426ee0d84..49d0c353084d71 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d_inbounds.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_2d_inbounds.mlir
@@ -1,24 +1,33 @@
-// RUN: mlir-opt %s --affine-super-vectorize="virtual-vector-size=4" | FileCheck %s
+// RUN: mlir-opt %s --affine-super-vectorize="virtual-vector-size=4" | FileCheck %s --check-prefix=MASKED
+// RUN: mlir-opt %s --affine-super-vectorize="virtual-vector-size=4" \
+// RUN: --eliminate-vector-masks --canonicalize | FileCheck %s
// RUN: mlir-opt %s \
// RUN: --convert-linalg-to-affine-loops \
// RUN: --affine-loop-tile="tile-sizes=16,16,16" \
// RUN: --affine-super-vectorize="virtual-vector-size=4" \
+// RUN: --eliminate-vector-masks \
// RUN: --canonicalize \
// RUN: | FileCheck %s --check-prefix=MATMUL
// RUN: mlir-opt %s \
// RUN: --affine-super-vectorize="virtual-vector-size=4" \
+// RUN: --eliminate-vector-masks --canonicalize \
// RUN: --convert-vector-to-llvm='enable-gep-inbounds-nuw=1' \
// RUN: --finalize-memref-to-llvm \
// RUN: --convert-func-to-llvm \
// RUN: | FileCheck %s --check-prefix=LLVM
+// The vectorizer masks each transfer by the memref bounds.
+// MASKED-LABEL: func.func @copy
+// MASKED: vector.mask %{{.*}} { vector.transfer_read
+// MASKED: vector.mask %{{.*}} { vector.transfer_write
+
// CHECK-LABEL: func.func @copy
-// Verify that transfer_read and transfer_write carry {in_bounds = [true]} when
-// the memref is static and its dimension is divisible by the vector width.
-// CHECK: vector.transfer_read {{.*}} {in_bounds = [true]} : memref<512x512xf32>, vector<4xf32>
-// CHECK-NOT: vector.transfer_read
-// CHECK: vector.transfer_write {{.*}} {in_bounds = [true]} : vector<4xf32>, memref<512x512xf32>
-// CHECK-NOT: vector.transfer_write
+// When the memref is static and its dimension is divisible by the vector
+// width, the masks are proven all-true and removed.
+// CHECK: %[[READ:.*]] = vector.transfer_read {{.*}} {in_bounds = [true]} : memref<512x512xf32>, vector<4xf32>
+// CHECK-NEXT: vector.transfer_write %[[READ]], {{.*}} {in_bounds = [true]} : vector<4xf32>, memref<512x512xf32>{{$}}
+// CHECK-NOT: vector.transfer_read
+// CHECK-NOT: vector.transfer_write
// LLVM-LABEL: llvm.func @copy
// Verify that in_bounds lowers to plain llvm.load/store with inbounds|nuw GEP,
@@ -41,10 +50,15 @@ func.func @copy(%A: memref<512x512xf32>, %B: memref<512x512xf32>) {
// MATMUL-LABEL: func.func @matmul
-// Verify all three transfer_read ops carry in_bounds=[true] after the full
-// linalg-to-affine + tiling + vectorization pipeline.
-// MATMUL-COUNT-3: vector.transfer_read {{.*}} {in_bounds = [true]
-// MATMUL-NOT: vector.transfer_read
+// Verify that all transfers are unmasked and carry in_bounds=[true] after the
+// full linalg-to-affine + tiling + vectorization pipeline.
+// MATMUL: %[[A:.*]] = vector.transfer_read {{.*}} {in_bounds = [true], permutation_map = #{{.*}}} : memref<512x512xf32>, vector<4xf32>
+// MATMUL-NEXT: %[[B:.*]] = vector.transfer_read {{.*}} {in_bounds = [true]} : memref<512x512xf32>, vector<4xf32>
+// MATMUL-NEXT: %[[C:.*]] = vector.transfer_read {{.*}} {in_bounds = [true]} : memref<512x512xf32>, vector<4xf32>
+// MATMUL-NEXT: %[[MUL:.*]] = arith.mulf %[[A]], %[[B]] : vector<4xf32>
+// MATMUL-NEXT: %[[ADD:.*]] = arith.addf %[[C]], %[[MUL]] : vector<4xf32>
+// MATMUL-NEXT: vector.transfer_write %[[ADD]], {{.*}} {in_bounds = [true]} : vector<4xf32>, memref<512x512xf32>{{$}}
+// MATMUL-NOT: vector.transfer_read
func.func @matmul(%A: memref<512x512xf32>,
%B: memref<512x512xf32>,
%C: memref<512x512xf32>) {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_3d.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_3d.mlir
index 17ff2c38c4eb50..fca99476a102d3 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_3d.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_3d.mlir
@@ -12,7 +12,7 @@ func.func @vec3d(%A : memref<?x?x?xf32>) {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 32 {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 64 {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256 {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} : memref<?x?x?xf32>, vector<32x64x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} : memref<?x?x?xf32>, vector<32x64x256xf32> }
affine.for %t0 = 0 to %0 {
affine.for %t1 = 0 to %0 {
affine.for %i0 = 0 to %0 {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_affine_apply.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_affine_apply.mlir
index 08ebea37b159d9..81e6a651117e98 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_affine_apply.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_affine_apply.mlir
@@ -12,8 +12,8 @@ func.func @vec_affine_apply(%arg0: memref<8x12x16xf32>, %arg1: memref<8x24x48xf3
// CHECK-NEXT: %[[S0:.*]] = affine.apply #[[$MAP_ID0]](%[[ARG3]])
// CHECK-NEXT: %[[S1:.*]] = affine.apply #[[$MAP_ID1]](%[[ARG4]])
// CHECK-NEXT: %[[CST:.*]] = ub.poison : f32
-// CHECK-NEXT: %[[S2:.*]] = vector.transfer_read %[[ARG0]][%[[ARG2]], %[[S0]], %[[S1]]], %[[CST]] {in_bounds = [true]} : memref<8x12x16xf32>, vector<8xf32>
-// CHECK-NEXT: vector.transfer_write %[[S2]], %[[ARG1]][%[[ARG2]], %[[ARG3]], %[[ARG4]]] {in_bounds = [true]} : vector<8xf32>, memref<8x24x48xf32>
+// CHECK: %[[S2:.*]] = vector.mask %{{.*}} { vector.transfer_read %[[ARG0]][%[[ARG2]], %[[S0]], %[[S1]]], %[[CST]] {in_bounds = [true]} : memref<8x12x16xf32>, vector<8xf32> }
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[S2]], %[[ARG1]][%[[ARG2]], %[[ARG3]], %[[ARG4]]] {in_bounds = [true]} : vector<8xf32>, memref<8x24x48xf32> }
// CHECK-NEXT: }
// CHECK-NEXT: }
// CHECK-NEXT: }
@@ -43,8 +43,8 @@ func.func @vec_affine_apply_2(%arg0: memref<8x12x16xf32>, %arg1: memref<8x24x48x
// CHECK-NEXT: affine.for %[[ARG4:.*]] = 0 to 48 step 8 {
// CHECK-NEXT: %[[S0:.*]] = affine.apply #[[$MAP_ID2]](%[[ARG4]])
// CHECK-NEXT: %[[CST:.*]] = ub.poison : f32
-// CHECK-NEXT: %[[S1:.*]] = vector.transfer_read %[[ARG0]][%[[ARG2]], %[[ARG3]], %[[S0]]], %[[CST]] {in_bounds = [true]} : memref<8x12x16xf32>, vector<8xf32>
-// CHECK-NEXT: vector.transfer_write %[[S1]], %[[ARG1]][%[[ARG2]], %[[ARG3]], %[[ARG4]]] {in_bounds = [true]} : vector<8xf32>, memref<8x24x48xf32>
+// CHECK: %[[S1:.*]] = vector.mask %{{.*}} { vector.transfer_read %[[ARG0]][%[[ARG2]], %[[ARG3]], %[[S0]]], %[[CST]] {in_bounds = [true]} : memref<8x12x16xf32>, vector<8xf32> }
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[S1]], %[[ARG1]][%[[ARG2]], %[[ARG3]], %[[ARG4]]] {in_bounds = [true]} : vector<8xf32>, memref<8x24x48xf32> }
// CHECK-NEXT: }
// CHECK-NEXT: }
// CHECK-NEXT: }
@@ -142,7 +142,7 @@ func.func @affine_map_with_expr_2(%arg0: memref<8x12x16xf32>, %arg1: memref<8x24
// CHECK-NEXT: %[[S2:.*]] = affine.apply #[[$MAP_ID5]](%[[ARG3]], %[[ARG4]], %[[I0]])
// CHECK-NEXT: %[[CST:.*]] = ub.poison : f32
// CHECK-NEXT: %[[S3:.*]] = vector.transfer_read %[[ARG0]][%[[S0]], %[[S1]], %[[S2]]], %[[CST]] {in_bounds = [true], permutation_map = #[[$MAP_ID6]]} : memref<8x12x16xf32>, vector<8xf32>
-// CHECK-NEXT: vector.transfer_write %[[S3]], %[[ARG1]][%[[ARG3]], %[[ARG4]], %[[ARG5]]] {in_bounds = [true]} : vector<8xf32>, memref<8x24x48xf32>
+// CHECK: vector.mask %{{.*}} { vector.transfer_write %[[S3]], %[[ARG1]][%[[ARG3]], %[[ARG4]], %[[ARG5]]] {in_bounds = [true]} : vector<8xf32>, memref<8x24x48xf32> }
// CHECK-NEXT: }
// CHECK-NEXT: }
// CHECK-NEXT: }
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_masked_read_write.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_masked_read_write.mlir
new file mode 100644
index 00000000000000..626b67389398e5
--- /dev/null
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_masked_read_write.mlir
@@ -0,0 +1,55 @@
+// RUN: mlir-opt %s -affine-super-vectorize="virtual-vector-size=8" | FileCheck %s
+
+// Verify the masks of the vectorized reads and writes: each one enables what is
+// left of the memref from the transfer index.
+
+// CHECK-LABEL: func.func @shifted_read
+// CHECK: affine.for %[[IV:.*]] = 0 to 63 step 8
+// CHECK: %[[IDX:.*]] = affine.apply #{{.*}}(%[[IV]])
+// CHECK: %[[SIZE:.*]] = arith.subi %{{.*}}, %[[IDX]] : index
+// CHECK: %[[MASK:.*]] = vector.create_mask %[[SIZE]] : vector<8xi1>
+// CHECK: vector.mask %[[MASK]] { vector.transfer_read %{{.*}}[%[[IDX]]]
+func.func @shifted_read(%src: memref<64xf32>, %dst: memref<64xf32>) {
+ affine.for %i = 0 to 63 {
+ %v = affine.load %src[%i + 1] : memref<64xf32>
+ affine.store %v, %dst[%i] : memref<64xf32>
+ }
+ return
+}
+
+// CHECK-LABEL: func.func @shifted_write
+// CHECK: affine.for %[[IV:.*]] = 0 to 60 step 8
+// CHECK: %[[IDX:.*]] = affine.apply #{{.*}}(%[[IV]])
+// CHECK: %[[SIZE:.*]] = arith.subi %{{.*}}, %[[IDX]] : index
+// CHECK: %[[MASK:.*]] = vector.create_mask %[[SIZE]] : vector<8xi1>
+// CHECK: vector.mask %[[MASK]] { vector.transfer_write %{{.*}}, %{{.*}}[%[[IDX]]]
+func.func @shifted_write(%src: memref<64xf32>, %dst: memref<64xf32>) {
+ affine.for %i = 0 to 60 {
+ %v = affine.load %src[%i] : memref<64xf32>
+ affine.store %v, %dst[%i + 4] : memref<64xf32>
+ }
+ return
+}
+
+// The inner loop is vectorized: only the memref dim it runs over is masked.
+
+// CHECK-LABEL: func.func @memref_2d_vector_1d
+// CHECK: affine.for %[[I:.*]] = 0 to 7
+// CHECK: affine.for %[[J:.*]] = 0 to 100 step 8
+// CHECK: %[[C100:.*]] = arith.constant 100 : index
+// CHECK: %[[RSIZE:.*]] = arith.subi %[[C100]], %[[J]] : index
+// CHECK: %[[RMASK:.*]] = vector.create_mask %[[RSIZE]] : vector<8xi1>
+// CHECK: vector.mask %[[RMASK]] { vector.transfer_read %{{.*}}[%[[I]], %[[J]]]
+// CHECK: %[[C100_1:.*]] = arith.constant 100 : index
+// CHECK: %[[WSIZE:.*]] = arith.subi %[[C100_1]], %[[J]] : index
+// CHECK: %[[WMASK:.*]] = vector.create_mask %[[WSIZE]] : vector<8xi1>
+// CHECK: vector.mask %[[WMASK]] { vector.transfer_write %{{.*}}, %{{.*}}[%[[I]], %[[J]]]
+func.func @memref_2d_vector_1d(%src: memref<7x100xf32>, %dst: memref<7x100xf32>) {
+ affine.for %i = 0 to 7 {
+ affine.for %j = 0 to 100 {
+ %v = affine.load %src[%i, %j] : memref<7x100xf32>
+ affine.store %v, %dst[%i, %j] : memref<7x100xf32>
+ }
+ }
+ return
+}
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_2d.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_2d.mlir
index 6b8f03ba9c6b5f..5c952f46ac0ff8 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_2d.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_2d.mlir
@@ -13,7 +13,7 @@ func.func @vec2d(%A : memref<?x?x?xf32>) {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 32
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256
- // CHECK: {{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d0d2]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: {{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d0d2]]} : memref<?x?x?xf32>, vector<32x256xf32> }
affine.for %i0 = 0 to %M {
affine.for %i1 = 0 to %N {
affine.for %i2 = 0 to %P {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_transpose_2d.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_transpose_2d.mlir
index 05465d734d0b3e..62fc952a1b5615 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_transpose_2d.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_outer_loop_transpose_2d.mlir
@@ -25,7 +25,7 @@ func.func @vec2d(%A : memref<?x?x?xf32>) {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 32
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
- // CHECK: {{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: {{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32> }
affine.for %i3 = 0 to %M {
affine.for %i4 = 0 to %N {
affine.for %i5 = 0 to %P {
@@ -46,12 +46,12 @@ func.func @vec2d_imperfectly_nested(%A : memref<?x?x?xf32>) {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 32 {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256 {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32> }
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256 {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32> }
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d0]]} : memref<?x?x?xf32>, vector<32x256xf32> }
affine.for %i0 = 0 to %0 {
affine.for %i1 = 0 to %1 {
affine.for %i2 = 0 to %2 {
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_reduction.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_reduction.mlir
index 836e7b0c71f1ff..3e67f80342ba1e 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_reduction.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_reduction.mlir
@@ -19,7 +19,7 @@ func.func @vecdim_reduction(%in: memref<256x512xf32>, %out: memref<256xf32>) {
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
// CHECK: }
@@ -46,7 +46,7 @@ func.func @vecdim_reduction_minf(%in: memref<256x512xf32>, %out: memref<256xf32>
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vmax:.*]] = arith.constant dense<0x7F800000> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vmax]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[min:.*]] = arith.minimumf %[[red_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[min]] : vector<128xf32>
// CHECK: }
@@ -73,7 +73,7 @@ func.func @vecdim_reduction_maxf(%in: memref<256x512xf32>, %out: memref<256xf32>
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vmin:.*]] = arith.constant dense<0xFF800000> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vmin]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[max:.*]] = arith.maximumf %[[red_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[max]] : vector<128xf32>
// CHECK: }
@@ -100,7 +100,7 @@ func.func @vecdim_reduction_minsi(%in: memref<256x512xi32>, %out: memref<256xi32
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vmax:.*]] = arith.constant dense<2147483647> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vmax]]) -> (vector<128xi32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[min:.*]] = arith.minsi %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[min]] : vector<128xi32>
// CHECK: }
@@ -127,7 +127,7 @@ func.func @vecdim_reduction_maxsi(%in: memref<256x512xi32>, %out: memref<256xi32
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vmin:.*]] = arith.constant dense<-2147483648> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vmin]]) -> (vector<128xi32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[max:.*]] = arith.maxsi %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[max]] : vector<128xi32>
// CHECK: }
@@ -154,7 +154,7 @@ func.func @vecdim_reduction_minui(%in: memref<256x512xi32>, %out: memref<256xi32
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vmax:.*]] = arith.constant dense<-1> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vmax]]) -> (vector<128xi32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[min:.*]] = arith.minui %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[min]] : vector<128xi32>
// CHECK: }
@@ -181,7 +181,7 @@ func.func @vecdim_reduction_maxui(%in: memref<256x512xi32>, %out: memref<256xi32
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vmin:.*]] = arith.constant dense<0> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vmin]]) -> (vector<128xi32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[max:.*]] = arith.maxui %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[max]] : vector<128xi32>
// CHECK: }
@@ -208,7 +208,7 @@ func.func @vecdim_reduction_andi(%in: memref<256x512xi32>, %out: memref<256xi32>
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vallone:.*]] = arith.constant dense<-1> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vallone]]) -> (vector<128xi32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[and:.*]] = arith.andi %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[and]] : vector<128xi32>
// CHECK: }
@@ -235,7 +235,7 @@ func.func @vecdim_reduction_ori(%in: memref<256x512xi32>, %out: memref<256xi32>)
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vzero:.*]] = arith.constant dense<0> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xi32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[or:.*]] = arith.ori %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[or]] : vector<128xi32>
// CHECK: }
@@ -266,7 +266,7 @@ func.func @vecdim_reduction_xori(%in: memref<256x512xi32>, %out: memref<256xi32>
// CHECK: %[[vzero:.*]] = arith.constant dense<0> : vector<128xi32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xi32>) {
// CHECK: %[[poison:.*]] = ub.poison : i32
-// CHECK: %[[ld:.*]] = vector.transfer_read %[[input]]{{\[}}%{{.*}}, %{{.*}}], %[[poison]] {in_bounds = [true]} : memref<256x512xi32>, vector<128xi32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %[[input]]{{\[}}%{{.*}}, %{{.*}}], %[[poison]] {in_bounds = [true]} : memref<256x512xi32>, vector<128xi32> }
// CHECK: %[[xor:.*]] = arith.xori %[[red_iter]], %[[ld]] : vector<128xi32>
// CHECK: affine.yield %[[xor]] : vector<128xi32>
// CHECK: }
@@ -299,7 +299,7 @@ func.func @vecdim_reduction_minnumf(%in: memref<256x512xf32>, %out: memref<256xf
// CHECK: %[[vzero:.*]] = arith.constant dense<0x7FC00000> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
// CHECK: %[[poison:.*]] = ub.poison : f32
-// CHECK: %[[ld:.*]] = vector.transfer_read %[[input]]{{\[}}%{{.*}}, %{{.*}}], %[[poison]] {in_bounds = [true]} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %[[input]]{{\[}}%{{.*}}, %{{.*}}], %[[poison]] {in_bounds = [true]} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[min:.*]] = arith.minnumf %[[red_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[min]] : vector<128xf32>
// CHECK: }
@@ -333,7 +333,7 @@ func.func @vecdim_reduction_maxnumf(%in: memref<256x512xf32>, %out: memref<256xf
// CHECK: %[[vzero:.*]] = arith.constant dense<0xFFC00000> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
// CHECK: %[[poison:.*]] = ub.poison : f32
-// CHECK: %[[ld:.*]] = vector.transfer_read %[[input]]{{\[}}%{{.*}}, %{{.*}}], %[[poison]] {in_bounds = [true]} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %[[input]]{{\[}}%{{.*}}, %{{.*}}], %[[poison]] {in_bounds = [true]} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[max:.*]] = arith.maxnumf %[[red_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[max]] : vector<128xf32>
// CHECK: }
@@ -366,7 +366,7 @@ func.func @vecdim_reduction_comm(%in: memref<256x512xf32>, %out: memref<256xf32>
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[add:.*]] = arith.addf %[[ld]], %[[red_iter]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
// CHECK: }
@@ -398,7 +398,7 @@ func.func @vecdim_reduction_expsin(%in: memref<256x512xf32>, %out: memref<256xf3
// CHECK: affine.for %{{.*}} = 0 to 256 {
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[vred:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[sin:.*]] = math.sin %[[ld]]
// CHECK: %[[exp:.*]] = math.exp %[[sin]]
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[exp]] : vector<128xf32>
@@ -435,7 +435,7 @@ func.func @two_vecdim_reductions(%in: memref<256x512xf32>, %out_sum: memref<256x
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[vone:.*]] = arith.constant dense<1.000000e+00> : vector<128xf32>
// CHECK: %[[vred:.*]]:2 = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[part_sum:.*]] = %[[vzero]], %[[part_prod:.*]] = %[[vone]]) -> (vector<128xf32>, vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[add:.*]] = arith.addf %[[part_sum]], %[[ld]] : vector<128xf32>
// CHECK: %[[mul:.*]] = arith.mulf %[[part_prod]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[add]], %[[mul]] : vector<128xf32>, vector<128xf32>
@@ -476,7 +476,7 @@ func.func @two_vecdim_reductions_int(%in: memref<256x512xi64>, %out_sum: memref<
// CHECK: %[[vzero:.*]] = arith.constant dense<0> : vector<128xi64>
// CHECK: %[[vone:.*]] = arith.constant dense<1> : vector<128xi64>
// CHECK: %[[vred:.*]]:2 = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[part_sum:.*]] = %[[vzero]], %[[part_prod:.*]] = %[[vone]]) -> (vector<128xi64>, vector<128xi64>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xi64>, vector<128xi64>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xi64>, vector<128xi64> }
// CHECK: %[[add:.*]] = arith.addi %[[part_sum]], %[[ld]] : vector<128xi64>
// CHECK: %[[mul:.*]] = arith.muli %[[part_prod]], %[[ld]] : vector<128xi64>
// CHECK: affine.yield %[[add]], %[[mul]] : vector<128xi64>, vector<128xi64>
@@ -511,7 +511,7 @@ func.func @vecdim_reduction_nested(%in: memref<256x512xf32>, %out: memref<1xf32>
// CHECK: %[[outer_red:.*]] = affine.for %{{.*}} = 0 to 512 step 128 iter_args(%[[outer_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
// CHECK: %[[vzero:.*]] = arith.constant dense<0.000000e+00> : vector<128xf32>
// CHECK: %[[inner_red:.*]] = affine.for %{{.*}} = 0 to 256 iter_args(%[[inner_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[add:.*]] = arith.addf %[[inner_iter]], %[[ld]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
// CHECK: }
@@ -648,7 +648,7 @@ func.func @vecdim_reduction_masked(%in: memref<256x512xf32>, %out: memref<256xf3
// CHECK: %[[vred:.*]] = affine.for %[[iv:.*]] = 0 to 500 step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
// CHECK: %[[elems_left:.*]] = affine.apply #[[$map0]](%[[iv]])
// CHECK: %[[mask:.*]] = vector.create_mask %[[elems_left]] : vector<128xi1>
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[select:.*]] = arith.select %[[mask]], %[[ld]], %[[vzero]] : vector<128xi1>, vector<128xf32>
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[select]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
@@ -681,7 +681,7 @@ func.func @vecdim_reduction_masked_unknown_ub(%in: memref<256x512xf32>, %out: me
// CHECK: %[[vred:.*]] = affine.for %[[iv:.*]] = 0 to %[[bnd:.*]] step 128 iter_args(%[[red_iter:.*]] = %[[vzero]]) -> (vector<128xf32>) {
// CHECK: %[[elems_left:.*]] = affine.apply #[[$map1]](%[[iv]])[%[[bnd]]]
// CHECK: %[[mask:.*]] = vector.create_mask %[[elems_left]] : vector<128xi1>
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[select:.*]] = arith.select %[[mask]], %[[ld]], %[[vzero]] : vector<128xi1>, vector<128xf32>
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[select]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
@@ -693,7 +693,8 @@ func.func @vecdim_reduction_masked_unknown_ub(%in: memref<256x512xf32>, %out: me
// -----
// The lower bound is nonzero, but the number of iterations is divisible by the
-// vector size, so masking is not needed.
+// vector size, so the reduction needs no mask. The read is still masked by the
+// memref bounds.
func.func @vecdim_reduction_nonzero_lb(%in: memref<256x512xf32>, %out: memref<256xf32>) {
%cst = arith.constant 0.000000e+00 : f32
@@ -709,8 +710,10 @@ func.func @vecdim_reduction_nonzero_lb(%in: memref<256x512xf32>, %out: memref<25
}
// CHECK-LABEL: @vecdim_reduction_nonzero_lb
-// CHECK: %{{.*}} = affine.for %{{.*}} = 127 to 511 step 128 iter_args({{.*}}) -> (vector<128xf32>) {
-// CHECK-NOT: vector.create_mask
+// CHECK: %{{.*}} = affine.for %{{.*}} = 127 to 511 step 128 iter_args(%[[RED_ITER:.*]] = %{{.*}}) -> (vector<128xf32>) {
+// CHECK: %[[LD:.*]] = vector.mask %{{.*}} { vector.transfer_read
+// CHECK-NEXT: %[[ADD:.*]] = arith.addf %[[RED_ITER]], %[[LD]] : vector<128xf32>
+// CHECK-NEXT: affine.yield %[[ADD]] : vector<128xf32>
// -----
@@ -735,7 +738,7 @@ func.func @vecdim_reduction_masked_unknown_lb(%in: memref<256x512xf32>, %out: me
// CHECK: %{{.*}} = affine.for %[[iv:.*]] = %[[lb:.*]] to 512 step 128 iter_args(%[[red_iter:.*]] = {{.*}}) -> (vector<128xf32>) {
// CHECK: %[[elems_left:.*]] = affine.apply #[[$map2]](%[[iv]])
// CHECK: %[[mask:.*]] = vector.create_mask %[[elems_left]] : vector<128xi1>
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[select:.*]] = arith.select %[[mask]], %[[ld]], %[[vzero]] : vector<128xi1>, vector<128xf32>
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[select]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
@@ -765,7 +768,7 @@ func.func @vecdim_reduction_complex_ub(%in: memref<256x512xf32>, %out: memref<25
// CHECK: %[[ub:.*]] = affine.min #[[$map3]](%[[M]], %[[N]])
// CHECK: %[[elems_left:.*]] = affine.apply #[[$map3_sub]](%[[iv]])[%[[ub]]]
// CHECK: %[[mask:.*]] = vector.create_mask %[[elems_left]] : vector<128xi1>
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[select:.*]] = arith.select %[[mask]], %[[ld]], %[[vzero]] : vector<128xi1>, vector<128xf32>
// CHECK: %[[add:.*]] = arith.addf %[[red_iter]], %[[select]] : vector<128xf32>
// CHECK: affine.yield %[[add]] : vector<128xf32>
@@ -798,7 +801,7 @@ func.func @vecdim_two_reductions_masked(%in: memref<256x512xf32>, %out: memref<5
// CHECK: %{{.*}} = affine.for %[[iv:.*]] = 0 to 500 step 128 iter_args(%[[sum_iter:.*]] = {{.*}}, %[[esum_iter:.*]] = {{.*}}) -> (vector<128xf32>, vector<128xf32>) {
// CHECK: %[[elems_left:.*]] = affine.apply #[[$map4]](%[[iv]])
// CHECK: %[[mask:.*]] = vector.create_mask %[[elems_left]] : vector<128xi1>
-// CHECK: %[[ld:.*]] = vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32>
+// CHECK: %[[ld:.*]] = vector.mask %{{.*}} { vector.transfer_read %{{.*}} : memref<256x512xf32>, vector<128xf32> }
// CHECK: %[[exp:.*]] = math.exp %[[ld]] : vector<128xf32>
// CHECK: %[[select0:.*]] = arith.select %[[mask]], %[[ld]], %[[vzero0]] : vector<128xi1>, vector<128xf32>
// CHECK: %[[add:.*]] = arith.addf %[[sum_iter]], %[[select0]] : vector<128xf32>
diff --git a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_transpose_2d.mlir b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_transpose_2d.mlir
index f1662b78242ed0..542b812f28fc06 100644
--- a/mlir/test/Dialect/Affine/SuperVectorize/vectorize_transpose_2d.mlir
+++ b/mlir/test/Dialect/Affine/SuperVectorize/vectorize_transpose_2d.mlir
@@ -2,6 +2,7 @@
// Permutation maps used in vectorization.
// CHECK-DAG: #[[map_proj_d0d1d2_d2d1:map[0-9]*]] = affine_map<(d0, d1, d2) -> (d2, d1)>
+// CHECK-DAG: #[[$map_proj_d0d1_d1d0:map[0-9]*]] = affine_map<(d0, d1) -> (d1, d0)>
func.func @vec2d(%A : memref<?x?x?xf32>) {
%c0 = arith.constant 0 : index
@@ -25,7 +26,7 @@ func.func @vec2d(%A : memref<?x?x?xf32>) {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 32
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256
- // CHECK: {{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: {{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32> }
affine.for %i3 = 0 to %M {
affine.for %i4 = 0 to %N {
affine.for %i5 = 0 to %P {
@@ -46,12 +47,12 @@ func.func @vec2d_imperfectly_nested(%A : memref<?x?x?xf32>) {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 32 {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256 {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32> }
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} {
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256 {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32> }
// CHECK: affine.for %{{.*}} = 0 to %{{.*}} step 256 {
- // CHECK: %{{.*}} = vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32>
+ // CHECK: %{{.*}} = vector.mask %{{.*}} { vector.transfer_read %{{.*}}[%{{.*}}, %{{.*}}, %{{.*}}], %{{.*}} {in_bounds = [true, true], permutation_map = #[[map_proj_d0d1d2_d2d1]]} : memref<?x?x?xf32>, vector<32x256xf32> }
affine.for %i0 = 0 to %0 {
affine.for %i1 = 0 to %1 {
affine.for %i2 = 0 to %2 {
@@ -70,3 +71,21 @@ func.func @vec2d_imperfectly_nested(%A : memref<?x?x?xf32>) {
return
}
+// The map selects the memref dims out of order, but the mask dims are in
+// increasing order.
+
+// CHECK-LABEL: func @transposed_store
+// CHECK: affine.for %[[I:.*]] = 0 to 128 step 32
+// CHECK: affine.for %[[J:.*]] = 0 to 512 step 256
+// CHECK: %[[SIZE0:.*]] = arith.subi %{{.*}}, %[[J]] : index
+// CHECK: %[[SIZE1:.*]] = arith.subi %{{.*}}, %[[I]] : index
+// CHECK: %[[MASK:.*]] = vector.create_mask %[[SIZE0]], %[[SIZE1]] : vector<256x32xi1>
+// CHECK: vector.mask %[[MASK]] { vector.transfer_write %{{.*}}, %{{.*}}[%[[J]], %[[I]]] {{.*}}permutation_map = #[[$map_proj_d0d1_d1d0]]{{.*}} : vector<32x256xf32>
+func.func @transposed_store(%A: memref<512x128xf32>, %f: f32) {
+ affine.for %i = 0 to 128 {
+ affine.for %j = 0 to 512 {
+ affine.store %f, %A[%j, %i] : memref<512x128xf32>
+ }
+ }
+ return
+}
diff --git a/mlir/test/Dialect/Linalg/vectorization/insert-slice.mlir b/mlir/test/Dialect/Linalg/vectorization/insert-slice.mlir
index 48e9b8823a187f..f60b2a93b4025b 100644
--- a/mlir/test/Dialect/Linalg/vectorization/insert-slice.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization/insert-slice.mlir
@@ -350,7 +350,6 @@ func.func private @insert_slice_vector_matches_dest_non_zero_offset(
// CHECK: %[[C_0_2:.*]] = arith.constant 0 : index
// CHECK: %[[C_2:.*]] = arith.constant 2 : index
// CHECK: %[[C_4_1:.*]] = arith.constant 4 : index
-// CHECK: %[[C_8:.*]] = arith.constant 8 : index
// CHECK: %[[C_6_1:.*]] = arith.constant 6 : index
// CHECK: %[[MASK_WRITE:.*]] = vector.create_mask %[[C_4_1]], %[[C_6_1]] : vector<4x8xi1>
// CHECK: %[[RES:.*]] = vector.mask %[[MASK_WRITE]] { vector.transfer_write %[[READ]], %[[DEST]][%[[C_0_2]], %[[C_2]]] {in_bounds = [true, true]} : vector<4x8xi32>, tensor<4x8xi32> } : vector<4x8xi1> -> tensor<4x8xi32>
@@ -479,7 +478,6 @@ func.func private @insert_slice_leading_unit_dim_dropped_non_zero_offset(
// CHECK: %[[C_0_2:.*]] = arith.constant 0 : index
// CHECK: %[[C_6:.*]] = arith.constant 6 : index
// CHECK: %[[C_0_3:.*]] = arith.constant 0 : index
-// CHECK: %[[C_8:.*]] = arith.constant 8 : index
// CHECK: %[[C_2_2:.*]] = arith.constant 2 : index
// CHECK: %[[C_8_1:.*]] = arith.constant 8 : index
// CHECK: %[[MASK_WRITE:.*]] = vector.create_mask %[[C_2_2]], %[[C_8_1]] : vector<4x2xi1>
@@ -517,14 +515,11 @@ func.func private @insert_slice_non_leading_unit_dim_dropped_non_zero_offsets(
// CHECK: %[[C_0:.*]] = arith.constant 0 : index
// CHECK: %[[C_0_1:.*]] = arith.constant 0 : index
// CHECK: %[[READ:.*]] = vector.transfer_read %[[SRC]][%[[C_0_1]], %[[C_0_1]]], %[[PAD]] {in_bounds = [true, true]} : tensor<4x4xi32>, vector<4x4xi32>
-// CHECK: %[[C_2:.*]] = arith.constant 2 : index
+// CHECK-NEXT: %[[C_2:.*]] = arith.constant 2 : index
// CHECK: %[[C_6:.*]] = arith.constant 6 : index
// CHECK: %[[C_0_2:.*]] = arith.constant 0 : index
// CHECK: %[[RES:.*]] = vector.transfer_write %[[READ]], %[[DEST]][%[[C_2]], %[[C_6]], %[[C_0_2]]] {in_bounds = [true, true], permutation_map = #[[$MAP_D0_D2]]} : vector<4x4xi32>, tensor<8x8x8xi32>
-// CHECK: %[[C_8:.*]] = arith.constant 8 : index
-// CHECK: %[[C_6_1:.*]] = arith.constant 6 : index
-// CHECK: %[[C_8_1:.*]] = arith.constant 8 : index
-// CHECK: return %[[RES]] : tensor<8x8x8xi32>
+// CHECK-NEXT: return %[[RES]] : tensor<8x8x8xi32>
module attributes {transform.with_named_sequence} {
transform.named_sequence @__transform_main(%arg0: !transform.any_op {transform.readonly}) {
>From 8ec88b1ab7f9ca7122f825c3ac3d95da1b4fef48 Mon Sep 17 00:00:00 2001
From: Hung-Kuan Tseng <tseng.tim096 at gmail.com>
Date: Sun, 22 Nov 2026 00:09:45 +0800
Subject: [PATCH 27/35] [mlir][vector] Only fold ext ops from the same type
into vector.contract (#221298)
`FoldArithExtIntoContractionOp` checks that both contraction operands
are produced by the same kind of extension, but not that the extensions
start from the same element type. With `extsi` from i8 on one side and
from i16 on the other it rewrites the contraction to take the two
sources directly, and the result fails to verify:
```
'vector.contract' op failed to verify that lhs and rhs have same element type
```
The same happens with `extf` from f16 and bf16. Compare the source
element types and leave such contractions alone.
Test: an `extf` case in `fold-arith-extf-into-vector-contract.mlir` that
keeps the extensions.
---
.../Vector/Transforms/VectorTransforms.cpp | 8 ++++++
.../fold-arith-extf-into-vector-contract.mlir | 25 +++++++++++++++++++
2 files changed, 33 insertions(+)
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
index ef221a212fc448..91caf78b0e595c 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
@@ -1940,6 +1940,14 @@ struct FoldArithExtIntoContractionOp
"no defining op on contract operands");
}
+ // The contraction requires lhs and rhs to have the same element type, so
+ // the two extensions must also start from the same one.
+ if (getElementTypeOrSelf(lhsDefOp.getIn().getType()) !=
+ getElementTypeOrSelf(rhsDefOp.getIn().getType())) {
+ return rewriter.notifyMatchFailure(
+ contractOp, "lhs and rhs are extended from different element types");
+ }
+
rewriter.replaceOpWithNewOp<vector::ContractionOp>(
contractOp, lhsDefOp->getOperand(0), rhsDefOp->getOperand(0),
contractOp.getAcc(), contractOp.getIndexingMapsAttr(),
diff --git a/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir b/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
index 99c904e7d024cc..e0d7a31127fee1 100644
--- a/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
+++ b/mlir/test/Dialect/Vector/fold-arith-extf-into-vector-contract.mlir
@@ -70,3 +70,28 @@ func.func @fold_arith_extsi_into_contract(
%lhs_i32, %rhs_i32, %arg2 : vector<64x64xi32>, vector<64x64xi32> into vector<64x64xi32>
return %result : vector<64x64xi32>
}
+
+// -----
+
+// The contraction requires lhs and rhs to have the same element type, so the
+// extensions are only folded when they start from the same one.
+
+// CHECK-LABEL: func.func @negative_fold_arith_extf_from_different_types
+// CHECK-SAME: (%[[ARG0:.*]]: vector<64x64xf16>, %[[ARG1:.*]]: vector<64x64xbf16>, %[[ARG2:.*]]: vector<64x64xf32>)
+// CHECK: %[[LHS:.*]] = arith.extf %[[ARG0]] : vector<64x64xf16> to vector<64x64xf32>
+// CHECK: %[[RHS:.*]] = arith.extf %[[ARG1]] : vector<64x64xbf16> to vector<64x64xf32>
+// CHECK: vector.contract
+// CHECK-SAME: %[[LHS]], %[[RHS]], %[[ARG2]] : vector<64x64xf32>, vector<64x64xf32> into vector<64x64xf32>
+func.func @negative_fold_arith_extf_from_different_types(
+ %arg0: vector<64x64xf16>,
+ %arg1: vector<64x64xbf16>,
+ %arg2: vector<64x64xf32>) -> vector<64x64xf32> {
+ %lhs_f32 = arith.extf %arg0 : vector<64x64xf16> to vector<64x64xf32>
+ %rhs_f32 = arith.extf %arg1 : vector<64x64xbf16> to vector<64x64xf32>
+ %result = vector.contract {
+ indexing_maps = [affine_map<(d0, d1, d2) -> (d0, d2)>, affine_map<(d0, d1, d2) -> (d2, d1)>, affine_map<(d0, d1, d2) -> (d0, d1)>],
+ iterator_types = ["parallel", "parallel", "reduction"],
+ kind = #vector.kind<add>}
+ %lhs_f32, %rhs_f32, %arg2 : vector<64x64xf32>, vector<64x64xf32> into vector<64x64xf32>
+ return %result : vector<64x64xf32>
+}
>From 8d350fa9589fbfa0196d2b3465b19b06922fbd46 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=C3=96mer=20Sinan=20A=C4=9Facan?= <omer at osa1.net>
Date: Sat, 21 Nov 2026 17:24:06 +0100
Subject: [PATCH 28/35] [LoopSimplify] Pass TTI and AC to SimplifyCFG's
foldBranchToCommonDest (#244537)
---
llvm/lib/Transforms/Utils/LoopSimplify.cpp | 2 +-
.../Transforms/LoopSimplify/merge-exits.ll | 39 ++++++++-----------
2 files changed, 18 insertions(+), 23 deletions(-)
diff --git a/llvm/lib/Transforms/Utils/LoopSimplify.cpp b/llvm/lib/Transforms/Utils/LoopSimplify.cpp
index 2d7ad6195c5799..91f0c764df1397 100644
--- a/llvm/lib/Transforms/Utils/LoopSimplify.cpp
+++ b/llvm/lib/Transforms/Utils/LoopSimplify.cpp
@@ -677,7 +677,7 @@ static bool simplifyOneLoop(Loop *L, SmallVectorImpl<Loop *> &Worklist,
// The block has now been cleared of all instructions except for
// a comparison and a conditional branch. SimplifyCFG may be able
// to fold it now.
- if (!foldBranchToCommonDest(BI, /*DTU=*/nullptr, MSSAU))
+ if (!foldBranchToCommonDest(BI, /*DTU=*/nullptr, MSSAU, TTI, AC))
continue;
// Success. The block is now dead, so remove it from the loop,
diff --git a/llvm/test/Transforms/LoopSimplify/merge-exits.ll b/llvm/test/Transforms/LoopSimplify/merge-exits.ll
index ad09dbeca0ac59..493b5b85acd06d 100644
--- a/llvm/test/Transforms/LoopSimplify/merge-exits.ll
+++ b/llvm/test/Transforms/LoopSimplify/merge-exits.ll
@@ -84,33 +84,28 @@ define float @merge_branches_profile_metadata(ptr %pTmp1, ptr %peakWeight, i32 %
; CHECK-NEXT: entry:
; CHECK-NEXT: [[T0:%.*]] = load float, ptr [[PEAKWEIGHT:%.*]], align 4
; CHECK-NEXT: [[T11:%.*]] = add i32 [[BANDEDGEINDEX:%.*]], -1
-; CHECK-NEXT: [[T121:%.*]] = icmp sgt i32 [[T11]], 0
-; CHECK-NEXT: br i1 [[T121]], label [[BB_LR_PH:%.*]], label [[BB3:%.*]], !prof [[PROF0:![0-9]+]]
-; CHECK: bb.lr.ph:
-; CHECK-NEXT: [[TMP0:%.*]] = sext i32 [[T11]] to i64
-; CHECK-NEXT: br label [[BB:%.*]]
+; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[T11]], i32 0)
+; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[SMAX]] to i64
+; CHECK-NEXT: br label [[BB2:%.*]]
; CHECK: bb:
-; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[BB_LR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[BB]] ]
-; CHECK-NEXT: [[DISTERBHI_04:%.*]] = phi float [ 0.000000e+00, [[BB_LR_PH]] ], [ [[T4:%.*]], [[BB]] ]
-; CHECK-NEXT: [[PEAKCOUNT_02:%.*]] = phi float [ [[T0]], [[BB_LR_PH]] ], [ [[T9:%.*]], [[BB]] ]
-; CHECK-NEXT: [[T2:%.*]] = getelementptr [4 x i8], ptr [[PTMP1:%.*]], i64 [[INDVARS_IV]]
+; CHECK-NEXT: [[T2:%.*]] = getelementptr [4 x i8], ptr [[PTMP1:%.*]], i64 [[INDVARS_IV:%.*]]
; CHECK-NEXT: [[T3:%.*]] = load float, ptr [[T2]], align 4
-; CHECK-NEXT: [[T4]] = fadd float [[T3]], [[DISTERBHI_04]]
-; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
+; CHECK-NEXT: [[T4:%.*]] = fadd float [[T3]], [[DISTERBHI_03:%.*]]
+; CHECK-NEXT: [[INDVARS_IV_NEXT:%.*]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[T7:%.*]] = getelementptr [4 x i8], ptr [[PEAKWEIGHT]], i64 [[INDVARS_IV_NEXT]]
; CHECK-NEXT: [[T8:%.*]] = load float, ptr [[T7]], align 4
-; CHECK-NEXT: [[T9]] = fadd float [[T8]], [[PEAKCOUNT_02]]
-; CHECK-NEXT: [[T10:%.*]] = fcmp olt float [[T4]], 2.500000e+00
-; CHECK-NEXT: [[T12:%.*]] = icmp sgt i64 [[TMP0]], [[INDVARS_IV_NEXT]]
-; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[T10]], [[T12]]
-; CHECK-NEXT: br i1 [[OR_COND]], label [[BB]], label [[BB1_BB3_CRIT_EDGE:%.*]], !prof [[PROF1:![0-9]+]]
-; CHECK: bb1.bb3_crit_edge:
-; CHECK-NEXT: [[T4_LCSSA:%.*]] = phi float [ [[T4]], [[BB]] ]
-; CHECK-NEXT: [[T9_LCSSA:%.*]] = phi float [ [[T9]], [[BB]] ]
-; CHECK-NEXT: br label [[BB3]]
+; CHECK-NEXT: [[T9:%.*]] = fadd float [[T8]], [[PEAKCOUNT_01:%.*]]
+; CHECK-NEXT: [[T10:%.*]] = fcmp uge float [[T4]], 2.500000e+00
+; CHECK-NEXT: br i1 [[T10]], label [[BB3:%.*]], label [[BB2]], !prof [[PROF0:![0-9]+]]
+; CHECK: bb2:
+; CHECK-NEXT: [[INDVARS_IV]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT]], [[BB:%.*]] ]
+; CHECK-NEXT: [[DISTERBHI_03]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[T4]], [[BB]] ]
+; CHECK-NEXT: [[PEAKCOUNT_01]] = phi float [ [[T0]], [[ENTRY]] ], [ [[T9]], [[BB]] ]
+; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV]], [[WIDE_TRIP_COUNT]]
+; CHECK-NEXT: br i1 [[EXITCOND]], label [[BB]], label [[BB3]]
; CHECK: bb3:
-; CHECK-NEXT: [[PEAKCOUNT_0_LCSSA:%.*]] = phi float [ [[T9_LCSSA]], [[BB1_BB3_CRIT_EDGE]] ], [ [[T0]], [[ENTRY:%.*]] ]
-; CHECK-NEXT: [[DISTERBHI_0_LCSSA:%.*]] = phi float [ [[T4_LCSSA]], [[BB1_BB3_CRIT_EDGE]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-NEXT: [[PEAKCOUNT_0_LCSSA:%.*]] = phi float [ [[PEAKCOUNT_01]], [[BB2]] ], [ [[T9]], [[BB]] ]
+; CHECK-NEXT: [[DISTERBHI_0_LCSSA:%.*]] = phi float [ [[DISTERBHI_03]], [[BB2]] ], [ [[T4]], [[BB]] ]
; CHECK-NEXT: [[T13:%.*]] = fdiv float [[PEAKCOUNT_0_LCSSA]], [[DISTERBHI_0_LCSSA]]
; CHECK-NEXT: ret float [[T13]]
;
>From c65400d3e273898ebecd47ea26ccb9bd7b3f9321 Mon Sep 17 00:00:00 2001
From: Kai Nacke <kai.peter.nacke at ibm.com>
Date: Sat, 21 Nov 2026 11:27:14 -0500
Subject: [PATCH 29/35] [SystemZ] Use NORK/NOGRK for bitwise NOT on z15+
(#244815)
With the miscellaneous-extensions-3 facility, `xor x, -1` can be
selected as a single NORK (i32) or NOGRK (i64) with both source operands
equal. This replaces LR+XILF for i32 and LCGR+AGHI for i64: one
instruction instead of two, with a shorter encoding and no register
copy.
Update affected tests and add not-01.ll coverage for z15 and z14.
Assisted by: IBM Bob
---
llvm/lib/Target/SystemZ/SystemZInstrInfo.td | 9 +++++-
.../test/CodeGen/SystemZ/fold-masked-merge.ll | 14 ++++----
llvm/test/CodeGen/SystemZ/not-01.ll | 23 +++++++++++++
llvm/test/CodeGen/SystemZ/pdep-pext.ll | 32 +++++++------------
llvm/test/CodeGen/SystemZ/shift-17.ll | 4 +--
5 files changed, 51 insertions(+), 31 deletions(-)
diff --git a/llvm/lib/Target/SystemZ/SystemZInstrInfo.td b/llvm/lib/Target/SystemZ/SystemZInstrInfo.td
index bc9e43148d6229..6634e1b341086e 100644
--- a/llvm/lib/Target/SystemZ/SystemZInstrInfo.td
+++ b/llvm/lib/Target/SystemZ/SystemZInstrInfo.td
@@ -2708,10 +2708,17 @@ let isCodeGenOnly = 1, hasSideEffects = 1, IsInsnDirective = 1 in {
def : Pat<(and (xor GR64:$x, (i64 -1)), GR64:$y),
(XGR GR64:$y, (NGR GR64:$y, GR64:$x))>;
-// Use LCGR/AGHI for i64 xor with -1.
+// Use LCGR/AGHI for i64 xor with -1 (pre-z15, see NOGRK below).
def : Pat<(xor GR64:$x, (i64 -1)),
(AGHI (LCGR GR64:$x), (i64 -1))>;
+// Use NOR with both operands equal for xor with -1 (bitwise NOT).
+// Complexity is added so that this is preferred over the LCGR/AGHI pattern.
+let Predicates = [FeatureMiscellaneousExtensions3], AddedComplexity = 1 in {
+ def : Pat<(xor GR32:$x, (i32 -1)), (NORK GR32:$x, GR32:$x)>;
+ def : Pat<(xor GR64:$x, (i64 -1)), (NOGRK GR64:$x, GR64:$x)>;
+}
+
// Shift/rotate instructions only use the last 6 bits of the second operand
// register, so we can safely use NILL (16 fewer bits than NILF) to only AND the
// last 16 bits.
diff --git a/llvm/test/CodeGen/SystemZ/fold-masked-merge.ll b/llvm/test/CodeGen/SystemZ/fold-masked-merge.ll
index c014345507f695..69b70df8914182 100644
--- a/llvm/test/CodeGen/SystemZ/fold-masked-merge.ll
+++ b/llvm/test/CodeGen/SystemZ/fold-masked-merge.ll
@@ -77,8 +77,7 @@ define i64 @masked_merge3(i64 %a0, i64 %a1, i64 %a2) {
;
; MISC3-LABEL: masked_merge3:
; MISC3: # %bb.0:
-; MISC3-NEXT: lcgr %r0, %r2
-; MISC3-NEXT: aghi %r0, -1
+; MISC3-NEXT: nogrk %r0, %r2, %r2
; MISC3-NEXT: ncgrk %r0, %r0, %r4
; MISC3-NEXT: ncgrk %r2, %r2, %r3
; MISC3-NEXT: ogr %r2, %r0
@@ -236,12 +235,11 @@ define i32 @masked_merge_no_transform1(i32 %a0, i32 %a1, i32 %a2, ptr %p1) {
;
; MISC3-LABEL: masked_merge_no_transform1:
; MISC3: # %bb.0:
-; MISC3-NEXT: nrk %r0, %r2, %r3
-; MISC3-NEXT: ncrk %r1, %r4, %r2
-; MISC3-NEXT: xilf %r2, 4294967295
-; MISC3-NEXT: or %r0, %r1
-; MISC3-NEXT: st %r2, 0(%r5)
-; MISC3-NEXT: lr %r2, %r0
+; MISC3-NEXT: nr %r3, %r2
+; MISC3-NEXT: nork %r0, %r2, %r2
+; MISC3-NEXT: ncrk %r2, %r4, %r2
+; MISC3-NEXT: or %r2, %r3
+; MISC3-NEXT: st %r0, 0(%r5)
; MISC3-NEXT: br %r14
%and0 = and i32 %a0, %a1
%not = xor i32 %a0, -1
diff --git a/llvm/test/CodeGen/SystemZ/not-01.ll b/llvm/test/CodeGen/SystemZ/not-01.ll
index 62248d6a1eb842..308ff5c488451a 100644
--- a/llvm/test/CodeGen/SystemZ/not-01.ll
+++ b/llvm/test/CodeGen/SystemZ/not-01.ll
@@ -1,6 +1,7 @@
; Combined logical operations involving complement on z15
;
; RUN: llc -mcpu=z15 < %s -mtriple=s390x-linux-gnu | FileCheck %s
+; RUN: llc -mcpu=z14 < %s -mtriple=s390x-linux-gnu | FileCheck %s --check-prefix=Z14
; And-with-complement 32-bit.
define i32 @f1(i32 %dummy, i32 %a, i32 %b) {
@@ -150,3 +151,25 @@ define i64 @f14(i64 %a) {
ret i64 %ret
}
+; Bitwise NOT 32-bit.
+define i32 @f15(i32 %dummy, i32 %a) {
+; CHECK-LABEL: f15:
+; CHECK: nork %r2, %r3, %r3
+; CHECK: br %r14
+; Z14-LABEL: f15:
+; Z14: xilf %r2, 4294967295
+ %ret = xor i32 %a, -1
+ ret i32 %ret
+}
+
+; Bitwise NOT 64-bit.
+define i64 @f16(i64 %dummy, i64 %a) {
+; CHECK-LABEL: f16:
+; CHECK: nogrk %r2, %r3, %r3
+; CHECK: br %r14
+; Z14-LABEL: f16:
+; Z14: lcgr
+; Z14: aghi %r2, -1
+ %ret = xor i64 %a, -1
+ ret i64 %ret
+}
diff --git a/llvm/test/CodeGen/SystemZ/pdep-pext.ll b/llvm/test/CodeGen/SystemZ/pdep-pext.ll
index 7a6b9c2f766f31..eb046da582eb01 100644
--- a/llvm/test/CodeGen/SystemZ/pdep-pext.ll
+++ b/llvm/test/CodeGen/SystemZ/pdep-pext.ll
@@ -21,8 +21,7 @@ define i64 @pdep_i64(i64 %a, i64 %mask) {
; CHECK-BASE-NEXT: .cfi_offset %r13, -56
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lcgr %r0, %r3
-; CHECK-BASE-NEXT: aghi %r0, -1
+; CHECK-BASE-NEXT: nogrk %r0, %r3, %r3
; CHECK-BASE-NEXT: sllg %r1, %r0, 1
; CHECK-BASE-NEXT: lgr %r4, %r1
; CHECK-BASE-NEXT: rxsbg %r4, %r0, 0, 61, 2
@@ -182,8 +181,7 @@ define i32 @pdep_i32(i32 %a, i32 %mask) {
; CHECK-BASE-NEXT: .cfi_offset %r13, -56
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lr %r0, %r3
-; CHECK-BASE-NEXT: xilf %r0, 4294967295
+; CHECK-BASE-NEXT: nork %r0, %r3, %r3
; CHECK-BASE-NEXT: sllk %r1, %r0, 1
; CHECK-BASE-NEXT: lgr %r4, %r1
; CHECK-BASE-NEXT: rxsbg %r4, %r0, 32, 61, 2
@@ -307,8 +305,7 @@ define i16 @pdep_i16(i16 %a, i16 %mask) {
; CHECK-BASE-NEXT: .cfi_offset %r13, -56
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lr %r0, %r3
-; CHECK-BASE-NEXT: xilf %r0, 4294967295
+; CHECK-BASE-NEXT: nork %r0, %r3, %r3
; CHECK-BASE-NEXT: sllk %r1, %r0, 1
; CHECK-BASE-NEXT: lgr %r4, %r1
; CHECK-BASE-NEXT: rxsbg %r4, %r0, 32, 61, 2
@@ -409,8 +406,7 @@ define i8 @pdep_i8(i8 %a, i8 %mask) {
; CHECK-BASE-NEXT: stmg %r14, %r15, 112(%r15)
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lr %r0, %r3
-; CHECK-BASE-NEXT: xilf %r0, 4294967295
+; CHECK-BASE-NEXT: nork %r0, %r3, %r3
; CHECK-BASE-NEXT: sllk %r1, %r0, 1
; CHECK-BASE-NEXT: lgr %r4, %r1
; CHECK-BASE-NEXT: rxsbg %r4, %r0, 32, 61, 2
@@ -488,12 +484,11 @@ define i8 @pdep_i8_const_mask(i8 %a) {
define i64 @pext_i64(i64 %a, i64 %mask) {
; CHECK-BASE-LABEL: pext_i64:
; CHECK-BASE: # %bb.0:
-; CHECK-BASE-NEXT: lcgr %r0, %r3
-; CHECK-BASE-NEXT: aghi %r0, -1
+; CHECK-BASE-NEXT: nogrk %r0, %r3, %r3
; CHECK-BASE-NEXT: sllg %r1, %r0, 1
; CHECK-BASE-NEXT: lgr %r4, %r1
-; CHECK-BASE-NEXT: ngr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r4, %r0, 0, 61, 2
+; CHECK-BASE-NEXT: ngr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r4, %r4, 0, 61, 2
; CHECK-BASE-NEXT: rxsbg %r4, %r4, 0, 59, 4
; CHECK-BASE-NEXT: rxsbg %r4, %r4, 0, 55, 8
@@ -633,12 +628,11 @@ define i32 @pext_i32(i32 %a, i32 %mask) {
; CHECK-BASE-NEXT: stmg %r14, %r15, 112(%r15)
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lr %r0, %r3
-; CHECK-BASE-NEXT: xilf %r0, 4294967295
+; CHECK-BASE-NEXT: nork %r0, %r3, %r3
; CHECK-BASE-NEXT: sllk %r4, %r0, 1
; CHECK-BASE-NEXT: lgr %r5, %r4
-; CHECK-BASE-NEXT: nr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r5, %r0, 32, 61, 2
+; CHECK-BASE-NEXT: nr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 61, 2
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 59, 4
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 55, 8
@@ -756,12 +750,11 @@ define i16 @pext_i16(i16 %a, i16 %mask) {
; CHECK-BASE-NEXT: stmg %r14, %r15, 112(%r15)
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lr %r0, %r3
-; CHECK-BASE-NEXT: xilf %r0, 4294967295
+; CHECK-BASE-NEXT: nork %r0, %r3, %r3
; CHECK-BASE-NEXT: sllk %r4, %r0, 1
; CHECK-BASE-NEXT: lgr %r5, %r4
-; CHECK-BASE-NEXT: nr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r5, %r0, 32, 61, 2
+; CHECK-BASE-NEXT: nr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 61, 2
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 59, 4
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 55, 8
@@ -859,12 +852,11 @@ define i8 @pext_i8(i8 %a, i8 %mask) {
; CHECK-BASE-NEXT: stmg %r14, %r15, 112(%r15)
; CHECK-BASE-NEXT: .cfi_offset %r14, -48
; CHECK-BASE-NEXT: .cfi_offset %r15, -40
-; CHECK-BASE-NEXT: lr %r0, %r3
-; CHECK-BASE-NEXT: xilf %r0, 4294967295
+; CHECK-BASE-NEXT: nork %r0, %r3, %r3
; CHECK-BASE-NEXT: sllk %r4, %r0, 1
; CHECK-BASE-NEXT: lgr %r5, %r4
-; CHECK-BASE-NEXT: nr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r5, %r0, 32, 61, 2
+; CHECK-BASE-NEXT: nr %r2, %r3
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 61, 2
; CHECK-BASE-NEXT: rxsbg %r5, %r5, 32, 59, 4
; CHECK-BASE-NEXT: ncrk %r0, %r4, %r5
diff --git a/llvm/test/CodeGen/SystemZ/shift-17.ll b/llvm/test/CodeGen/SystemZ/shift-17.ll
index 8f5f9abd0540b5..031a60684cb9b2 100644
--- a/llvm/test/CodeGen/SystemZ/shift-17.ll
+++ b/llvm/test/CodeGen/SystemZ/shift-17.ll
@@ -115,7 +115,7 @@ define i128 @f4(i128 %a, i128 %b, i128 %sh) {
; Z15-NEXT: vslb %v1, %v1, %v2
; Z15-NEXT: vsl %v1, %v1, %v2
; Z15-NEXT: vrepib %v2, 1
-; Z15-NEXT: xilf %r0, 4294967295
+; Z15-NEXT: nork %r0, %r0, %r0
; Z15-NEXT: vsrl %v0, %v0, %v2
; Z15-NEXT: vlvgp %v2, %r0, %r0
; Z15-NEXT: vrepb %v2, %v2, 15
@@ -236,7 +236,7 @@ define i128 @f8(i128 %a, i128 %b, i128 %sh) {
; Z15-NEXT: vsrlb %v1, %v1, %v2
; Z15-NEXT: vsrl %v1, %v1, %v2
; Z15-NEXT: vrepib %v2, 1
-; Z15-NEXT: xilf %r0, 4294967295
+; Z15-NEXT: nork %r0, %r0, %r0
; Z15-NEXT: vsl %v0, %v0, %v2
; Z15-NEXT: vlvgp %v2, %r0, %r0
; Z15-NEXT: vrepb %v2, %v2, 15
>From 4c01a2991df8237f0eeada0c32ed5c3807aa9707 Mon Sep 17 00:00:00 2001
From: Dhairyashil R G <dhairyashil25 at gmail.com>
Date: Sat, 21 Nov 2026 22:00:23 +0530
Subject: [PATCH 30/35] [mlir][linalg] Do not fold tensor.pad into a read
inside vector.mask (#243927)
`PadOpVectorizationWithTransferReadPattern` folds a `tensor.pad` into a
`vector.transfer_read` of its result and replaces the read's padding
with the pad value. It bails out on a mask operand, but not on
`vector.mask`, where the masked-off lanes take the read's padding.
For example, vectorizing a `linalg.pack` (tile 8, padding `-1`) of a
`tensor.pad` (3 to 5 elements, value `7`) gives a masked read of the
pad. After `pad_vectorization`, the packed row is `1 2 3 7 7 7 7 7`
instead of `1 2 3 7 7 -1 -1 -1`.
The pattern now bails out on reads inside `vector.mask` too. Tested by
`@negative_pad_and_transfer_read_in_vector_mask`.
Part of #227821 and #239511.
Assisted-by: Claude
---
.../Linalg/Transforms/Vectorization.cpp | 4 +--
.../Linalg/vectorization/pad-patterns.mlir | 35 +++++++++++++++++++
2 files changed, 37 insertions(+), 2 deletions(-)
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
index 3e9761825ef5c5..2557215e0b3096 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
@@ -2761,7 +2761,7 @@ struct VectorizePadOpUserPattern : public OpRewritePattern<tensor::PadOp> {
/// sure that the original padding value %cst was never used.
///
/// This rewrite is possible if:
-/// - `xferOp` has no out-of-bounds dims or mask.
+/// - `xferOp` has no out-of-bounds dims and is not masked.
/// - Low padding is static 0.
/// - Single, scalar padding value.
struct PadOpVectorizationWithTransferReadPattern
@@ -2779,7 +2779,7 @@ struct PadOpVectorizationWithTransferReadPattern
if (!padValue)
return failure();
// Padding value of existing `xferOp` is unused.
- if (xferOp.hasOutOfBoundsDim() || xferOp.getMask())
+ if (xferOp.hasOutOfBoundsDim() || xferOp.getMask() || xferOp.isMasked())
return failure();
rewriter.modifyOpInPlace(xferOp, [&]() {
diff --git a/mlir/test/Dialect/Linalg/vectorization/pad-patterns.mlir b/mlir/test/Dialect/Linalg/vectorization/pad-patterns.mlir
index 65c30e052f4b2d..42981a410e0dee 100644
--- a/mlir/test/Dialect/Linalg/vectorization/pad-patterns.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization/pad-patterns.mlir
@@ -46,6 +46,41 @@ module attributes {transform.with_named_sequence} {
// -----
+// The masked-off lanes take %c6, so the pad can't be folded into the read.
+
+// CHECK-LABEL: func @negative_pad_and_transfer_read_in_vector_mask
+// CHECK-DAG: %[[C6:.*]] = arith.constant 6.0
+// CHECK: %[[PAD:.*]] = tensor.pad
+// CHECK: vector.mask %{{.*}} { vector.transfer_read %[[PAD]][%{{.*}}, %{{.*}}], %[[C6]]
+func.func @negative_pad_and_transfer_read_in_vector_mask(
+ %arg0: tensor<5x6xf32>, %mask: vector<7x9xi1>) -> vector<7x9xf32> {
+ %c0 = arith.constant 0 : index
+ %c5 = arith.constant 5.0 : f32
+ %c6 = arith.constant 6.0 : f32
+ %0 = tensor.pad %arg0 low[0, 0] high[5, 7] {
+ ^bb0(%arg1: index, %arg2: index):
+ tensor.yield %c5 : f32
+ } : tensor<5x6xf32> to tensor<10x13xf32>
+ %1 = vector.mask %mask {
+ vector.transfer_read %0[%c0, %c0], %c6
+ : tensor<10x13xf32>, vector<7x9xf32>
+ } : vector<7x9xi1> -> vector<7x9xf32>
+ return %1 : vector<7x9xf32>
+}
+
+module attributes {transform.with_named_sequence} {
+ transform.named_sequence @__transform_main(%arg1: !transform.any_op {transform.readonly}) {
+ %func_op = transform.structured.match ops{["func.func"]} in %arg1 : (!transform.any_op) -> !transform.op<"func.func">
+
+ transform.apply_patterns to %func_op {
+ transform.apply_patterns.linalg.pad_vectorization
+ } : !transform.op<"func.func">
+ transform.yield
+ }
+}
+
+// -----
+
///----------------------------------------------------------------------------------------
/// [Pattern: PadOpVectorizationWithTransferWritePattern]
///----------------------------------------------------------------------------------------
>From d9ec1f2c2645601cd4621609666b45d6bd04f7eb Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev at outlook.com>
Date: Sat, 21 Nov 2026 11:42:56 -0500
Subject: [PATCH 31/35] Revert "[AArch64] Share materialized constant between
add x,k and sub x,k (#225430)"
This reverts commit aabb9609da40d6257a94c991d16ed96f6a81f0bf to fix
miscompilation, reported in https://github.com/llvm/llvm-project/pull/225430#issuecomment-7039795522
Pull-request: https://github.com/llvm/llvm-project/pull/244907
---
.../Target/AArch64/AArch64MIPeepholeOpt.cpp | 115 +-----------------
llvm/test/CodeGen/AArch64/pr222095.ll | 105 ----------------
2 files changed, 4 insertions(+), 216 deletions(-)
delete mode 100644 llvm/test/CodeGen/AArch64/pr222095.ll
diff --git a/llvm/lib/Target/AArch64/AArch64MIPeepholeOpt.cpp b/llvm/lib/Target/AArch64/AArch64MIPeepholeOpt.cpp
index 32c97413c11fb7..f47d19cf3a05a1 100644
--- a/llvm/lib/Target/AArch64/AArch64MIPeepholeOpt.cpp
+++ b/llvm/lib/Target/AArch64/AArch64MIPeepholeOpt.cpp
@@ -72,8 +72,6 @@
#include "AArch64ExpandImm.h"
#include "AArch64InstrInfo.h"
#include "MCTargetDesc/AArch64AddressingModes.h"
-#include "llvm/ADT/DenseMap.h"
-#include "llvm/ADT/STLExtras.h"
#include "llvm/CodeGen/MachineDominators.h"
#include "llvm/CodeGen/MachineLoopInfo.h"
@@ -124,15 +122,6 @@ class AArch64MIPeepholeOptImpl {
bool checkMovImmInstr(MachineInstr &MI, MachineInstr *&MovMI,
MachineInstr *&SubregToRegMI);
- bool getConstantDef(Register Reg, uint64_t &Value, MachineInstr *&MovMI,
- MachineInstr *&SubregToRegMI);
- template <typename T>
- void recordConstantDef(MachineInstr &MI,
- DenseMap<uint64_t, Register> &Consts);
- bool foldSharedAddSubConstant(MachineInstr &MI, unsigned AddOpc,
- unsigned SubOpc, bool Is64Bit,
- DenseMap<uint64_t, Register> &SameWidth);
-
template <typename T>
bool visitADDSUB(unsigned PosOpc, unsigned NegOpc, MachineInstr &MI);
template <typename T>
@@ -590,82 +579,6 @@ bool AArch64MIPeepholeOptImpl::checkMovImmInstr(MachineInstr &MI,
return true;
}
-bool AArch64MIPeepholeOptImpl::getConstantDef(Register Reg, uint64_t &Value,
- MachineInstr *&MovMI,
- MachineInstr *&SubregToRegMI) {
- SubregToRegMI = nullptr;
- MovMI = MRI->getUniqueVRegDef(Reg);
- if (!MovMI)
- return false;
- if (MovMI->getOpcode() == TargetOpcode::SUBREG_TO_REG) {
- SubregToRegMI = MovMI;
- MovMI = MRI->getUniqueVRegDef(MovMI->getOperand(1).getReg());
- if (!MovMI)
- return false;
- }
- if (MovMI->getOpcode() != AArch64::MOVi32imm &&
- MovMI->getOpcode() != AArch64::MOVi64imm)
- return false;
- Value = static_cast<uint64_t>(MovMI->getOperand(1).getImm());
- if (SubregToRegMI)
- Value = static_cast<uint32_t>(Value);
- return true;
-}
-
-static bool hasRealUse(const MachineRegisterInfo &MRI, Register Reg) {
- return any_of(MRI.use_nodbg_instructions(Reg),
- [](const MachineInstr &UseMI) { return !UseMI.isFakeUse(); });
-}
-
-template <typename T>
-void AArch64MIPeepholeOptImpl::recordConstantDef(
- MachineInstr &MI, DenseMap<uint64_t, Register> &Consts) {
- if (hasRealUse(*MRI, MI.getOperand(0).getReg()))
- Consts[static_cast<T>(MI.getOperand(1).getImm())] =
- MI.getOperand(0).getReg();
-}
-
-bool AArch64MIPeepholeOptImpl::foldSharedAddSubConstant(
- MachineInstr &MI, unsigned AddOpc, unsigned SubOpc, bool Is64Bit,
- DenseMap<uint64_t, Register> &SameWidth) {
- unsigned Opc = MI.getOpcode();
-
- Register SrcReg = MI.getOperand(2).getReg();
- if (!SrcReg.isVirtual())
- return false;
-
- uint64_t RawValue;
- MachineInstr *MovMI, *SubregToRegMI;
- if (!getConstantDef(SrcReg, RawValue, MovMI, SubregToRegMI))
- return false;
-
- uint64_t NegValue =
- Is64Bit ? (0ULL - RawValue)
- : static_cast<uint32_t>(0U - static_cast<uint32_t>(RawValue));
-
- auto SameIt = SameWidth.find(NegValue);
- if (SameIt == SameWidth.end() || SameIt->second == SrcReg ||
- MRI->def_empty(SameIt->second))
- return false;
- Register CanonReg = SameIt->second;
-
- unsigned NewOpc = (Opc == AddOpc) ? SubOpc : AddOpc;
- MI.getOperand(2).setReg(CanonReg);
- MI.setDesc(TII->get(NewOpc));
- MRI->clearKillFlags(CanonReg);
-
- if (MRI->use_nodbg_empty(SrcReg)) {
- if (SubregToRegMI) {
- SubregToRegMI->eraseFromParent();
- if (MRI->use_nodbg_empty(MovMI->getOperand(0).getReg()))
- MovMI->eraseFromParent();
- } else {
- MovMI->eraseFromParent();
- }
- }
- return true;
-}
-
template <typename T>
bool AArch64MIPeepholeOptImpl::splitTwoPartImm(MachineInstr &MI,
SplitAndOpcFunc<T> SplitAndOpc,
@@ -1073,18 +986,10 @@ bool AArch64MIPeepholeOptImpl::run(MachineFunction &MF) {
bool Changed = false;
for (MachineBasicBlock &MBB : MF) {
- DenseMap<uint64_t, Register> SharedConsts32, SharedConsts64;
-
for (MachineInstr &MI : make_early_inc_range(MBB)) {
switch (MI.getOpcode()) {
default:
break;
- case AArch64::MOVi32imm:
- recordConstantDef<uint32_t>(MI, SharedConsts32);
- break;
- case AArch64::MOVi64imm:
- recordConstantDef<uint64_t>(MI, SharedConsts64);
- break;
case AArch64::INSERT_SUBREG:
Changed |= visitINSERT(MI);
break;
@@ -1124,28 +1029,16 @@ bool AArch64MIPeepholeOptImpl::run(MachineFunction &MF) {
Changed |= visitORR(MI);
break;
case AArch64::ADDWrr:
- Changed |=
- foldSharedAddSubConstant(MI, AArch64::ADDWrr, AArch64::SUBWrr,
- false, SharedConsts32) ||
- visitADDSUB<uint32_t>(AArch64::ADDWri, AArch64::SUBWri, MI);
+ Changed |= visitADDSUB<uint32_t>(AArch64::ADDWri, AArch64::SUBWri, MI);
break;
case AArch64::SUBWrr:
- Changed |=
- foldSharedAddSubConstant(MI, AArch64::ADDWrr, AArch64::SUBWrr,
- false, SharedConsts32) ||
- visitADDSUB<uint32_t>(AArch64::SUBWri, AArch64::ADDWri, MI);
+ Changed |= visitADDSUB<uint32_t>(AArch64::SUBWri, AArch64::ADDWri, MI);
break;
case AArch64::ADDXrr:
- Changed |=
- foldSharedAddSubConstant(MI, AArch64::ADDXrr, AArch64::SUBXrr, true,
- SharedConsts64) ||
- visitADDSUB<uint64_t>(AArch64::ADDXri, AArch64::SUBXri, MI);
+ Changed |= visitADDSUB<uint64_t>(AArch64::ADDXri, AArch64::SUBXri, MI);
break;
case AArch64::SUBXrr:
- Changed |=
- foldSharedAddSubConstant(MI, AArch64::ADDXrr, AArch64::SUBXrr, true,
- SharedConsts64) ||
- visitADDSUB<uint64_t>(AArch64::SUBXri, AArch64::ADDXri, MI);
+ Changed |= visitADDSUB<uint64_t>(AArch64::SUBXri, AArch64::ADDXri, MI);
break;
case AArch64::ADDSWrr:
Changed |=
diff --git a/llvm/test/CodeGen/AArch64/pr222095.ll b/llvm/test/CodeGen/AArch64/pr222095.ll
deleted file mode 100644
index 3c75e234cd5fba..00000000000000
--- a/llvm/test/CodeGen/AArch64/pr222095.ll
+++ /dev/null
@@ -1,105 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
-; RUN: llc -mtriple=aarch64-linux-gnu < %s | FileCheck %s
-
-; https://github.com/llvm/llvm-project/issues/222095
-; When the same constant magnitude is needed once added and once subtracted,
-; and that magnitude doesn't fit as a direct immediate on this target, the
-; positive form should be materialized once and shared (via a plain `sub`)
-; rather than also materializing an unrelated negated constant.
-
-define i32 @add_sub_shared_const_i32(i32 %a, i32 %b) {
-; CHECK-LABEL: add_sub_shared_const_i32:
-; CHECK: // %bb.0:
-; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
-; CHECK-NEXT: .cfi_def_cfa_offset 16
-; CHECK-NEXT: .cfi_offset w30, -16
-; CHECK-NEXT: mov w8, #-4097 // =0xffffefff
-; CHECK-NEXT: sub w0, w0, w8
-; CHECK-NEXT: add w1, w1, w8
-; CHECK-NEXT: bl use32
-; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
-; CHECK-NEXT: ret
- %add = add i32 %a, 4097
- %sub = add i32 %b, -4097
- %r = call i32 @use32(i32 %add, i32 %sub)
- ret i32 %r
-}
-declare i32 @use32(i32, i32)
-
-define i64 @add_sub_shared_const_i64(i64 %a, i64 %b) {
-; CHECK-LABEL: add_sub_shared_const_i64:
-; CHECK: // %bb.0:
-; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
-; CHECK-NEXT: .cfi_def_cfa_offset 16
-; CHECK-NEXT: .cfi_offset w30, -16
-; CHECK-NEXT: mov x8, #-4097 // =0xffffffffffffefff
-; CHECK-NEXT: sub x0, x0, x8
-; CHECK-NEXT: add x1, x1, x8
-; CHECK-NEXT: bl use64
-; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
-; CHECK-NEXT: ret
- %add = add i64 %a, 4097
- %sub = add i64 %b, -4097
- %r = call i64 @use64(i64 %add, i64 %sub)
- ret i64 %r
-}
-declare i64 @use64(i64, i64)
-
-; Three or more shared uses in the same block, mixing add and sub, should all
-; converge on one materialized register.
-define i32 @add_sub_shared_const_multi_use(i32 %a, i32 %b, i32 %c) {
-; CHECK-LABEL: add_sub_shared_const_multi_use:
-; CHECK: // %bb.0:
-; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
-; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
-; CHECK-NEXT: .cfi_def_cfa_offset 32
-; CHECK-NEXT: .cfi_offset w19, -8
-; CHECK-NEXT: .cfi_offset w20, -16
-; CHECK-NEXT: .cfi_offset w30, -32
-; CHECK-NEXT: mov w20, #-4097 // =0xffffefff
-; CHECK-NEXT: mov w19, w2
-; CHECK-NEXT: sub w0, w0, w20
-; CHECK-NEXT: add w1, w1, w20
-; CHECK-NEXT: bl use32
-; CHECK-NEXT: add w1, w19, w20
-; CHECK-NEXT: bl use32
-; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
-; CHECK-NEXT: ret
- %add = add i32 %a, 4097
- %sub1 = add i32 %b, -4097
- %sub2 = add i32 %c, -4097
- %r1 = call i32 @use32(i32 %add, i32 %sub1)
- %r2 = call i32 @use32(i32 %r1, i32 %sub2)
- ret i32 %r2
-}
-
-; The same tie-break must not fire when the other operand is a multiply:
-; madd can fold the constant directly into its accumulator, which is
-; strictly better than materializing the positive magnitude and using a
-; separate sub. This is a regression guard for that specific interaction.
-define i64 @add_neg_tied_const_does_not_break_madd(i64 %a) {
-; CHECK-LABEL: add_neg_tied_const_does_not_break_madd:
-; CHECK: // %bb.0:
-; CHECK-NEXT: mov w8, #37 // =0x25
-; CHECK-NEXT: mov x9, #-32888 // =0xffffffffffff7f88
-; CHECK-NEXT: movk x9, #65518, lsl #16
-; CHECK-NEXT: madd x0, x0, x8, x9
-; CHECK-NEXT: ret
- %tmp0 = add i64 %a, -31000
- %tmp1 = mul i64 %tmp0, 37
- ret i64 %tmp1
-}
-
-; Self-negating edge case: for i32, -2147483648 (INT_MIN) negated wraps back
-; to itself, so there is no positive magnitude to converge on. Must not
-; needlessly flip the opcode for no benefit.
-define i32 @add_self_negating_const(i32 %a) {
-; CHECK-LABEL: add_self_negating_const:
-; CHECK: // %bb.0:
-; CHECK-NEXT: mov w8, #-2147483648 // =0x80000000
-; CHECK-NEXT: add w0, w0, w8
-; CHECK-NEXT: ret
- %add = add i32 %a, -2147483648
- ret i32 %add
-}
>From d1d4bfa045a5f6fe99b496f05f2231d6c7943a3f Mon Sep 17 00:00:00 2001
From: Kunal Dubey <xakep8 at protonmail.com>
Date: Sat, 21 Nov 2026 22:41:19 +0530
Subject: [PATCH 32/35] [Clang] add relese note for 232629 (#244864)
The release note for was not added for #232629.
---
clang/docs/ReleaseNotes.md | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/clang/docs/ReleaseNotes.md b/clang/docs/ReleaseNotes.md
index 4f4cfaf7af73a5..63e17e4880a8db 100644
--- a/clang/docs/ReleaseNotes.md
+++ b/clang/docs/ReleaseNotes.md
@@ -84,6 +84,10 @@ in a future version of Clang.
### ABI Changes in This Version
+- Clang now correctly mangles ``new std::initializer_list<T>{...}``
+ expressions under the Itanium C++ ABI. Previous versions emitted a duplicate
+ initializer-list marker in the mangled name. (#GH232531)
+
- Clang now correctly mangles C++20 parenthesized aggregate initialization
expressions. Previously, initializer arguments could be omitted from the
mangled name, causing distinct entities to receive the same name.
>From c93f7fabe831aafabd158831175ab123c9e77ec6 Mon Sep 17 00:00:00 2001
From: Raul Tambre <raul at tambre.ee>
Date: Sat, 21 Nov 2026 19:15:26 +0200
Subject: [PATCH 33/35] [cmake][runtimes] Pass variable type for passthrough
CMake options (#227662)
Passing the types affects the handling of `PATH` cache variables.
The motivating case is setting `COMPILER_RT_INSTALL_LIBRARY_DIR` without `CMAKE_BINARY_DIR` getting prepended to it.
Previously reverted because of https://github.com/llvm/llvm-project/pull/198505#issuecomment-4496493320.
However there's insufficient information to replicate or fix the non-upstream build issue so I'll go ahead with reapplying in the hopes that maybe it's magically resolved itself.
Revert-Of: 313ab4dc3b3a5f4c14079c82806c1f28975f724f
Re-Apply-Of: 2e06e008a7b44fab5c0d7fbc4331c696b5e19b06
---
llvm/runtimes/CMakeLists.txt | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/llvm/runtimes/CMakeLists.txt b/llvm/runtimes/CMakeLists.txt
index 0401ab00c6ecc5..9fba87fb7a5868 100644
--- a/llvm/runtimes/CMakeLists.txt
+++ b/llvm/runtimes/CMakeLists.txt
@@ -107,6 +107,14 @@ macro(append_passthrough_options out type names)
list(APPEND ${out} -C ${cache})
endforeach()
else()
+ # Type is important for cache variables.
+ # For example PATH variables will be made relative to CMAKE_BINARY_DIR
+ # at the point of set(CACHE) if the type isn't passed along.
+ get_property(type CACHE ${variable_name} PROPERTY TYPE)
+ if(NOT type STREQUAL "UNINITIALIZED")
+ string(APPEND new_name ":${type}")
+ endif()
+
string(REPLACE ";" "|" new_value "${${variable_name}}")
list(APPEND ${out} "-D${new_name}=${new_value}")
endif()
>From 90044d5e4534506fdc3be5fea87274dda5930c74 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=C3=96mer=20Sinan=20A=C4=9Facan?= <omer at osa1.net>
Date: Sat, 21 Nov 2026 08:31:13 +0000
Subject: [PATCH 34/35] Add test
---
.../unnamed-functions.ll | 52 +++++++++++++++++++
1 file changed, 52 insertions(+)
create mode 100644 llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll
diff --git a/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll b/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll
new file mode 100644
index 00000000000000..c0a5c3454c5acb
--- /dev/null
+++ b/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll
@@ -0,0 +1,52 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -passes=called-value-propagation -S < %s | FileCheck %s
+
+; Check handling of unnamed functions.
+
+define internal void @0() {
+; CHECK-LABEL: define internal void @0() {
+; CHECK-NEXT: ret void
+;
+ ret void
+}
+
+define internal void @1() {
+; CHECK-LABEL: define internal void @1() {
+; CHECK-NEXT: ret void
+;
+ ret void
+}
+
+define void @named_a() {
+; CHECK-LABEL: define void @named_a() {
+; CHECK-NEXT: ret void
+;
+ ret void
+}
+
+define void @named_b() {
+; CHECK-LABEL: define void @named_b() {
+; CHECK-NEXT: ret void
+;
+ ret void
+}
+
+define void @test(i1 %c) {
+; CHECK-LABEL: define void @test(
+; CHECK-SAME: i1 [[C:%.*]]) {
+; CHECK-NEXT: [[P:%.*]] = select i1 [[C]], ptr @[[GLOB0:[0-9]+]], ptr @[[GLOB1:[0-9]+]]
+; CHECK-NEXT: call void [[P]](), !callees [[META0:![0-9]+]]
+; CHECK-NEXT: [[Q:%.*]] = select i1 [[C]], ptr @named_a, ptr @named_b
+; CHECK-NEXT: call void [[Q]](), !callees [[META1:![0-9]+]]
+; CHECK-NEXT: ret void
+;
+ %p = select i1 %c, ptr @0, ptr @1
+ call void %p()
+ %q = select i1 %c, ptr @named_a, ptr @named_b
+ call void %q()
+ ret void
+}
+;.
+; CHECK: [[META0]] = !{ptr @[[GLOB0]]}
+; CHECK: [[META1]] = !{ptr @named_a, ptr @named_b}
+;.
>From 03cd85e1832d1e89f22f334f4610eb1dd63cd9ae Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=C3=96mer=20Sinan=20A=C4=9Facan?= <omer at osa1.net>
Date: Sat, 21 Nov 2026 08:31:46 +0000
Subject: [PATCH 35/35] Fix it
---
.../Transforms/IPO/CalledValuePropagation.cpp | 34 +++++++++++++------
.../unnamed-functions.ll | 2 +-
2 files changed, 24 insertions(+), 12 deletions(-)
diff --git a/llvm/lib/Transforms/IPO/CalledValuePropagation.cpp b/llvm/lib/Transforms/IPO/CalledValuePropagation.cpp
index 4bfd374f357964..ecfd73ff1eae23 100644
--- a/llvm/lib/Transforms/IPO/CalledValuePropagation.cpp
+++ b/llvm/lib/Transforms/IPO/CalledValuePropagation.cpp
@@ -50,19 +50,21 @@ class CVPLatticeVal {
/// refer.
enum CVPLatticeStateTy { Undefined, FunctionSet, Overdefined, Untracked };
- /// Comparator for sorting the functions set. We want to keep the order
- /// deterministic for testing, etc.
+ /// Comparator for sorting the functions set. Compare using the unique and
+ /// deterministic function indices for deterministic output.
struct Compare {
+ const DenseMap<const Function *, uint32_t> &FunctionIndices;
bool operator()(const Function *LHS, const Function *RHS) const {
- return LHS->getName() < RHS->getName();
+ return FunctionIndices.at(LHS) < FunctionIndices.at(RHS);
}
};
CVPLatticeVal() = default;
CVPLatticeVal(CVPLatticeStateTy LatticeState) : LatticeState(LatticeState) {}
- CVPLatticeVal(std::vector<Function *> &&Functions)
+ CVPLatticeVal(std::vector<Function *> &&Functions,
+ const DenseMap<const Function *, uint32_t> &FunctionIndices)
: LatticeState(FunctionSet), Functions(std::move(Functions)) {
- assert(llvm::is_sorted(this->Functions, Compare()));
+ assert(llvm::is_sorted(this->Functions, Compare{FunctionIndices}));
}
/// Get a reference to the functions held by this lattice value. The number
@@ -104,13 +106,16 @@ class CVPLatticeVal {
class CVPLatticeFunc
: public AbstractLatticeFunction<CVPLatticeKey, CVPLatticeVal> {
const IPOOptions &Opts;
+ const DenseMap<const Function *, uint32_t> &FunctionIndices;
public:
- explicit CVPLatticeFunc(const IPOOptions &Opts)
+ explicit CVPLatticeFunc(
+ const IPOOptions &Opts,
+ const DenseMap<const Function *, uint32_t> &FunctionIndices)
: AbstractLatticeFunction(CVPLatticeVal(CVPLatticeVal::Undefined),
CVPLatticeVal(CVPLatticeVal::Overdefined),
CVPLatticeVal(CVPLatticeVal::Untracked)),
- Opts(Opts) {}
+ Opts(Opts), FunctionIndices(FunctionIndices) {}
/// Compute and return a CVPLatticeVal for the given CVPLatticeKey.
CVPLatticeVal ComputeLatticeVal(CVPLatticeKey Key) override {
@@ -150,10 +155,11 @@ class CVPLatticeFunc
std::vector<Function *> Union;
std::set_union(X.getFunctions().begin(), X.getFunctions().end(),
Y.getFunctions().begin(), Y.getFunctions().end(),
- std::back_inserter(Union), CVPLatticeVal::Compare{});
+ std::back_inserter(Union),
+ CVPLatticeVal::Compare{FunctionIndices});
if (Union.size() > Opts.cvp_max_functions_per_value)
return getOverdefinedVal();
- return CVPLatticeVal(std::move(Union));
+ return CVPLatticeVal(std::move(Union), FunctionIndices);
}
/// Compute the lattice values that change as a result of executing the given
@@ -225,7 +231,7 @@ class CVPLatticeFunc
if (isa<ConstantPointerNull>(C))
return CVPLatticeVal(CVPLatticeVal::FunctionSet);
if (auto *F = dyn_cast<Function>(C->stripPointerCasts()))
- return CVPLatticeVal({F});
+ return CVPLatticeVal({F}, FunctionIndices);
return getOverdefinedVal();
}
@@ -365,8 +371,14 @@ template <> struct LatticeKeyInfo<CVPLatticeKey> {
} // namespace llvm
static bool runCVP(Module &M) {
+ // Assign unique numbers to functions, to be used when ordering the functions
+ // in the lattice.
+ DenseMap<const Function *, uint32_t> FunctionIndices;
+ for (const auto &[Idx, F] : enumerate(M))
+ FunctionIndices.try_emplace(&F, Idx);
+
// Our custom lattice function and generic sparse propagation solver.
- CVPLatticeFunc Lattice(getIPOOptions(M));
+ CVPLatticeFunc Lattice(getIPOOptions(M), FunctionIndices);
SparseSolver<CVPLatticeKey, CVPLatticeVal> Solver(&Lattice);
// For each function in the module, if we can't track its arguments, let the
diff --git a/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll b/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll
index c0a5c3454c5acb..75f66910788782 100644
--- a/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll
+++ b/llvm/test/Transforms/CalledValuePropagation/unnamed-functions.ll
@@ -47,6 +47,6 @@ define void @test(i1 %c) {
ret void
}
;.
-; CHECK: [[META0]] = !{ptr @[[GLOB0]]}
+; CHECK: [[META0]] = !{ptr @[[GLOB0]], ptr @[[GLOB1]]}
; CHECK: [[META1]] = !{ptr @named_a, ptr @named_b}
;.
More information about the llvm-branch-commits
mailing list